Mechanisms of QT prolongation by buprenorphine cannot be explained by direct hERG channel block

Buprenorphine is a μ-opioid receptor (MOR) partial agonist used to manage pain and addiction. QTC prolongation that crosses the 10 msec threshold of regulatory concern was observed at a supratherapeutic dose in two thorough QT studies for the transdermal buprenorphine product BUTRANS®. Because QTC prolongation can be associated with Torsades de Pointes (TdP), a rare but potentially fatal ventricular arrhythmia, these results have led to further investigation of the electrophysiological effects of buprenorphine. Drug-induced QTC prolongation and TdP are most commonly caused by acute inhibition of hERG current (IhERG) that contribute to the repolarizing phase of the ventricular action potentials (APs). Concomitant inhibition of inward late Na+ (INaL) and/or L-type Ca2+ (ICaL) current can offer some protection against proarrhythmia. Therefore, we characterized the effects of buprenorphine and its major metabolite norbuprenorphine on cardiac hERG, Ca2+, and Na+ ion channels, as well as cardiac APs. For comparison, methadone, a MOR agonist associated with QTC prolongation and high TdP risk, and naltrexone and naloxone, two opioid receptor antagonists, were also studied. Whole cell recordings were performed at 37°C on cells stably expressing hERG, CaV1.2, and NaV1.5 proteins. Microelectrode array (MEA) recordings were made on human induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs). The results showed that buprenorphine, norbuprenorphine, naltrexone, and naloxone had no effect on IhERG, ICaL, INaL, and peak Na+ current (INaP) at clinically relevant concentrations. In contrast, methadone inhibited IhERG, ICaL, and INaL. Experiments on iPSC-CMs showed a lack of effect for buprenorphine, norbuprenorphine, naltrexone, and naloxone, and delayed repolarization for methadone at clinically relevant concentrations. The mechanism of QTC prolongation is opioid moiety-specific. This remains undefined for buprenorphine, while for methadone it involves direct hERG channel block. There is no evidence that buprenorphine use is associated with TdP. Whether this lack of TdP risk can be generalized to other drugs with QTC prolongation not mediated by acute hERG channel block warrants further study.


Introduction
Buprenorphine is a μ-opioid receptor (MOR) partial agonist used to manage opioid use disorder and pain severe enough to warrant use of opioid analgesics. In healthy volunteers, multi-day exposure to a transdermal formulation BUTRANS 1 has been linked with dose-dependent QT C prolongation that reaches the threshold of regulatory concern (10 ms) for the supratherapeutic dose [1]. This effect was not observed when buprenorphine was co-administered with naltrexone, an opioid receptor antagonist. Because there are buprenorphine products that achieve higher systemic exposure levels at recommended doses than the supratherapeutic dose in the BUTRANS 1 study, and because QT C prolongation can be associated with a rare but potentially fatal ventricular tachyarrhythmia called Torsades de Pointes (TdP), these results have led to questions about whether buprenorphine's QTc prolongation is associated with any risk of TdP.
The most common cause of drug-induced QT C prolongation and TdP is acute inhibition of hERG current (I hERG ) that is important for the repolarizing phase of the ventricular action potential (AP) [2,3]. Examples of drugs that prolong the QT C interval yet are associated with low TdP risk (i.e., ranolazine, amiodarone, and verapamil) suggest that concomitant inhibition of late Na + (I NaL ) [4] and/or L-type Ca 2+ current (I CaL ) [5,6] could offset the TdP risk imposed by hERG channel block. Though less common, some drugs prolong the QT C interval by increasing I NaL [7,8]. Basic science studies also show that delayed repolarization and ventricular arrhythmias can occur by pharmacologically enhancing I CaL [9]. Knowledge of drug effects on inward and outward ventricular ionic currents is thus important for proarrhythmia risk assessment. Buprenorphine has been reported to inhibit I hERG [10] and peak Na + current (I NaP ) [11] in the micromolar to tens of micromolar range. These concentrations far exceed the subnanomolar free drug level achieved by therapeutic dosing. However, the studies were performed at room temperature. Given that potencies of some hERG channel blockers are temperature-sensitive [12][13][14][15][16][17], it is possible that buprenorphine's potencies against I hERG and I NaP would be different if experiments were conducted at a physiological temperature. The hERG data mentioned above were obtained using hERG1a-overexpressing cells [10]. In human ventricular myocytes, hERG channels are formed by heteromeric assembly of hERG1a and hERG1b subunits [18]. Some drugs exhibit different potencies in blocking homomeric (hERG1a) and heteromeric hERG channels [19], and buprenorphine's effect on heteromeric hERG channels has not been assessed. Additionally, a literature search produced no information regarding buprenorphine's effects on I CaL and I NaL , nor multi-ion channel pharmacology data for norbuprenorphine, the major active metabolite of buprenorphine, that is also a MOR partial agonist.
To address the abovementioned knowledge gap, we performed patch clamp studies at 37˚C to characterize the effects of buprenorphine and norbuprenorphine on I hERG mediated by hERG1a and hERG1a/1b channels, I CaL , I NaL , and I NaP using overexpression cell lines, and microelectrode array (MEA) experiments to study drug effects on the electrical behavior of human induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs). For comparison, methadone, a MOR receptor full agonist that is associated with QT C prolongation and high TdP risk, and two opioid receptor antagonists, naltrexone and naloxone, were also studied.
For data transparency, the original patch clamp records or individual data points required to construct graphs presented in the figures are available for download at: https://osf.io/tjuev/.
The detailed cell culture procedures for overexpression cell lines can be found at: https:// osf.io/tjuev/.

Whole cell voltage clamp electrophysiology
All experiments were conducted at 37˚C. Glass coverslips with attached cells were placed in a recording chamber mounted on an inverted (Zeiss Axiovert 135TV) or an upright microscope (Zeiss AxioExaminer D1). The recording chamber was continuously perfused with external solution flowing at a rate of 2-3 mL/min. Recordings were made using borosilicate glass pipettes (BF150-86-10; Sutter Instrument, CA) pulled with a micropipette puller (P97; Sutter Instrument, CA) to 1.5-3 MO resistance when filled with internal solution.
For I hERG , the extracellular solution contained (in mM): 130 NaCl, 10 HEPES, 5 KCl, 1 MgCl 2 �6H 2 O, 1 CaCl 2 �2H 2 O, and 12.5 dextrose; pH was adjusted to 7.4 with 5 M NaOH. The internal solution contained (in mM): 120 K-gluconate, 20 KCl, 10 HEPES, 5 EGTA, and 1.5 MgATP; pH was adjusted to 7.3 with 1 M KOH;~280 mOsM. The voltage command values were corrected for the 15 mV liquid junction potential (calculated using pClamp10 software; Molecular Device, CA) that resulted from using these solutions. Cells were depolarized from a holding potential of -80 mV to +40 mV for 2 s, then repolarized to -50 mV for 1.5 s to evoke the tail current that is predominantly mediated by hERG channels, and finally returned to -80 mV. A 100 ms hyperpolarizing step from -80 mV to -90 mV was included prior to the depolarizing step to monitor input resistance throughout the recording. This voltage protocol was repeated at 15 s intervals.
For I CaL , the external solution contained (in mM): 137 NaCl, 4 KCl, 1.8 CaCl 2 , 1 MgCl 2 , 10 HEPES, and 10 dextrose; pH was adjusted to 7.4 with 5M NaOH. The intracellular solution contained (in mM): 120 aspartic acid, 120 CsOH, 10 CsCl, 10 EGTA, 5 MgATP, 0.4 TrisGTP, and 10 HEPES; pH was adjusted to 7.2 with 5M CsOH;~290 mOsM. The voltage command values were corrected for the 17 mV liquid junction potential. Cells were held at -80 mV, depolarized to 0 mV for 40 ms, further depolarized to +30 mV for 200 ms, and then ramped down to -80 mV in 100 ms (-1.1 V/s). A 100 ms hyperpolarizing step from -80 to -90 mV was included prior to the first depolarizing step to monitor input resistance throughout the recording. This voltage protocol was presented at 5 s intervals.
For I NaP and I NaL recordings, the external solution contained (in mM): 130 NaCl, 4 CsCl, 2 CaCl 2 , 1 MgCl 2 , 10 HEPES, and 10 dextrose; pH was adjusted to 7.4 with NaOH. The internal solution contained (in mM): 130 CsCl, 7 NaCl, 1 MgCl 2 , 5 EGTA, and 5 HEPES; pH was adjusted to 7.2 with CsOH;~280 mOsM. When recording I NaL , 150 nM ATX-II was added to the external solution to slow Na + channel inactivation, thereby inducing a pronounced late component. Cells were first hyperpolarized from -95 mV to -120 mV for 200 ms to facilitate recovery of Na + channels from inactivation, then depolarized to -15 mV for 40 ms, then +40 mV for 200 ms, then ramped down to -95 mV in 100 ms (-1.35 V/s). This voltage protocol was repeated at 12.5 s intervals.
These voltage protocols were continuously presented at the respective intervals throughout the duration of the recordings. Recordings were obtained first in control solution until the amplitude of the ionic current being studied reached stability, then drug solution was bathapplied and recording repeated in the presence of drug until a new steady state current amplitude was reached. For I CaL , current rundown upon whole cell formation exhibited multiple phases. Stability for I CaL experiments was thus defined as rundown reaching a steady and slower rate. Each cell was exposed to one or more concentrations of the same drug as long as membrane properties (i.e., input resistance and holding current at rest) and recording quality (current signal) remained stable.
Cells were visualized using phase contrast method for the inverted microscope and differential interference contrast-infrared (DIC-IR) method for the upright microscope. Temperatures of the in-line solution heater and recording chamber were maintained with a dual channel temperature controller (TC2BIP from Cell MicroControls for the inverted microscope setup; TC-344C from Warner Instruments for the upright microscope setup), and temperature of the perfusate near the cells was recorded throughout the experiment with a thermistor. Recordings were obtained using Multiclamp 700B amplifiers (Molecular Devices, CA). For I hERG , signals were filtered at 2.2 kHz and digitized at 5 kHz using a Digidata 1550A interface (Molecular Devices, CA), and transferred to a computer using pClamp10 software (Molecular Devices, CA). For I CaL , signals were filtered at 3 kHz and digitized at 10 kHz. For I NaP and I NaL , signals were filtered at 10 kHz and digitized at 20 kHz. Seal resistance was always >1 GO, and series resistance was electronically compensated at 80%. To allow for adequate internal solution dialysis prior to actual current recording, following whole cell formation cells were given~2 min resting period, during which the recording and membrane properties were monitored using the "membrane test" function of the pClamp10 software.

Data and statistical analysis for whole cell voltage clamp experiments
Current traces were analyzed using custom macros written in Igor Pro 6 (WaveMetrics) and pClamp10 software (Molecular Devices, CA). For I hERG and I NaP , the peak current amplitude was subtracted against baseline current amplitude measured at the beginning of the voltage protocol (-80 mV for I hERG and -95 mV for I NaP experiments). For I CaL and I NaL , the current amplitude for each recorded trace was subtracted against calculated passive current for that trace as previously described [22]. Passive current calculation was done by first using Ohm's Law to calculate input resistance using the current evoked by the -10 mV hyperpolarizing step (from -80 mV to -90 mV) for the I CaL voltage protocol or the -25 mV hyperpolarizing step (from -95 mV to -120 mV) for the I NaL voltage protocol, and then using the input resistance value and Ohm's law to calculate the passive current expected at various voltages. In a subset of Ca V 1.2 cells, despite the use of Cs + -based internal solution, an endogenous outward current was revealed following I CaL inhibition. For these cells, traces obtained in the presence of 100 μM verapamil (with endogenous current present) were used for subtraction for all recorded traces to obtain I CaL amplitude. In some cells tested with 100 to 300 μM methadone, verapamil was not applied and complete abolition of I CaL was achieved. When endogenous outward current was found for these cells, traces obtained in methadone were used for subtraction to calculate I CaL amplitude.
The potencies of tested drugs against various ionic currents were quantified by constructing concentration-inhibition plots. For I hERG recordings, fractional inhibition from individual cells was calculated by first normalizing the average current amplitude of the last 3 consecutively recorded traces in drug solution to that obtained in control solution, and then subtracting this value from unity. For I CaL recordings, the averaged current amplitude of 10 consecutively recorded traces in drug and control solutions were used. For I NaL and I NaP recordings, the averaged values of 5 consecutively recorded traces in drug and control solutions were used. For each current, the fractional inhibition values from all cells were then pooled and plotted against drug concentrations to generate concentration-inhibition plots, and these plots were fit with the Hill equation using Igor Pro 6 (WaveMetrics, Portland, OR, USA), with minimum and maximum fractional block constrained to 0 and 1, respectively. Half inhibitory concentration (IC 50 ) and the Hill coefficient (n H ) obtained from the fit are presented as mean ± standard deviation (SD) in the text. Summary data points in the figures are shown as mean ± standard error of the mean (SEM). Statistical analysis was performed on the concentration-inhibition plots for buprenorphine on hERG1a and hERG1a/1b channels using GraphPad Prism version 8.3. Individual datapoints for each concentration were fitted to the Hill equation with variable slope, in the form of "Fractional block = 1 / (1+ (IC 50 /[buprenorphine]) nH )", and the best-fit values for the IC 50 and n H were compared using the extra-sumof-squares F test with the significance P value set to < 0.05. These data and statistical analysis plans comply with the recommendations on experimental design and analysis in pharmacology [23].

Cell culture for human iPSC-CMs and MEA plating
iCell cardiomyocytes 2 were purchased from Fujifilm Cellular Dynamics (cat. # R1017; FCDI, Madison, Wisconsin). These cells were thawed, centrifuged at 180 g, and seeded at 50,000 cells/well in 5 μL droplets on the MEA 48-well plates (cat. # M768-KAP-48, Axion Biosystems). Just prior to use, the MEA plates were coated with fibronectin (1:20 with DPBS; cat. # 11080938001,1mg/mL, Roche) for 1 hr at 37˚C. After 1 hr, 300 μL of maintenance media was slowly added to each well in the plate. Cells on MEA plates were maintained at 37˚C and 5% CO 2 according to published procedures [24].

MEA recordings
All drug testing was performed on day 7 of plating cells on the MEA plates. One day prior to the experiment, iPSC-CMs were fed with fresh maintenance media. On the day of the experiment, 1000X stock solution in DMSO was prepared for each tested drug concentration. These 1000X stocks were then made into 10X stocks by diluting with maintenance media. Final concentration of DMSO did not exceed 0.1%. DMSO (0.1%) and vehicle controls were randomly distributed across the plate to normalize for edge-effects on the multi-well MEA Maestro system (Axion Biosystems). Separate drug-dosing plates were prepared and equilibrated at 37˚C and 5% CO 2 . MEA plates equilibrated on the Maestro for 20 min before baseline recordings. The drugs were then added by pipetting 30 μL of the 10X stock solution into each well containing 270 μL of the medium. Post-drug data were collected 30 min after drug application. Field potentials in iPSC-CMs were recorded using AxIS software version 2.4.2 (Axion Biosystems) in cardiac standard configuration (130X gain) with a sampling frequency of 12.5 kHz, and a band pass filter of 0.1-2000 Hz. A statistics compiler tool filtered the beats with spike amplitude (SA) > 0.3 mV, beat-to-beat field potential duration (FPD) consistency within 2X SD, electrode FPD consistency (10% coefficient of variation), and well FPD consistency within 2X median absolute deviation. Thirty stable beats from the last 5 min of each recording were selected for analysis using Axis CiPA analysis tool (version 1.2.3).

Data analysis for MEA recordings
The following data inclusion criteria were applied to the baseline parameters: 1) iPSC-CM baseline spontaneous beating rate within 20-90 beats per minute (0.3-1.5 Hz); 2) baseline beating rate within 6 SDs calculated from the baseline beating rate for all wells on a given plate; 3) coefficient of variation for the baseline beat period (BP) less than 5%; and 4) SA > 0.3 mV. SA is defined as the peak-to-peak amplitude of the depolarization spike. These inclusion criteria were adopted from a published study [25].
FPD of the spontaneously-beating iPSC-CMs were corrected using Fridericia's correction formula, , where BP is iPSC-CM beat period (averaged time between consecutive depolarization spikes in MEA recordings). Double delta FPD (ΔΔFPD), double-delta beat period (ΔΔBP), and double delta spike amplitude (ΔΔSA) were calculated after correcting drug-induced FPD with plate-specific vehicle (as DMSO control) and well-based baseline FPD values. ΔΔFPD data were excluded from analysis if more than 50% of wells for a given drug concentration exhibited arrhythmia-like events.

Drugs
The following drugs and solvent were purchased from Sigma-Aldrich: buprenorphine hydrochloride (B9275, USDEA C-III), (±)-methadone hydrochloride (M0267, USDEA C-II), naltrexone hydrochloride (N3136), and DMSO (D8418). Naloxone hydrochloride (0599), (±)-verapamil hydrochloride (0654), E-4031 dihydrochloride (1808), and tetrodotoxin citrate or TTX (1069) were purchased from Tocris Bioscience. Norbuprenorphine hydrochloride was purchased from Noramco. ATX-II (STA-700) was purchased from Alomone Labs. To make stock solutions for patch clamp experiments, TTX, verapamil, E-4031, ATX-II, naloxone, and naltrexone were dissolved in water. Buprenorphine, norbuprenorphine, and methadone were dissolved in DMSO. When DMSO was used as a solvent to prepare stock solution, the final DMSO concentration applied to overexpression cells was �0.3% for Ca V 1.2 cells and �0.1% for the rest. In iPSC-CM experiments, stock solutions were all made with DMSO, and the final DMSO concentration applied to the cells was �0.1%. Aliquoted stock solutions were stored at -20˚C until the day of experimentation and were thawed, vortexed, and diluted to specific test concentrations in extracellular solution.
Clinically relevant concentrations of the tested drugs. The IC 50 s obtained for individual drugs were compared with the maximum systemic concentrations of free or total drug following therapeutic dosing (free or total C max , respectively) to assess the likelihood of acute cardiac ion channel block when these drugs are used as indicated. The preference for this comparison is free C max if the percent of plasma protein binding is known, since free drug molecules are thought to be the ones that could interact with cardiac ion channels. C max values used in this paper reflect the highest across all drug products for a specific moiety to reflect the highest therapeutic exposure level. For buprenorphine, free C max was calculated using data from the drug label for Sublocade™. Following the 4 th monthly dose of subcutaneous injection, the steady state plasma buprenorphine level reached 10.12 ng/mL, or 0.02 μM converted using a molecular weight of 467.64 g/mol. Given 96% plasma protein binding, free C max for buprenorphine was 0.00087 μM. For norbuprenorphine, total C max was based on the data after the 7 th sublingual, 24 mg dose of Subutex™ (https://www.ema.europa.eu/en/documents/assessment-report/ buvidal-epar-public-assessment-report_en.pdf). Total C max was 9.29 ng/mL or 0.022 μM, converted using a molecular weight of 413.55 g/mol. Free C max was not calculated as the percent of plasma protein binding for norbuprenorphine has not been empirically determined. In-house simulation using GastroPlus package (Simulation Plus Inc.) predicted this metabolite to be extensively plasma protein-bound (95 ± 5%), just like the parent molecule. For methadone, free C max was calculated using a total C max of 904 ng/mL from Table 4 of the study by Florian and colleagues [26]. Given 85% plasma protein binding, free C max of methadone was 135.6 ng/ mL or 0.44 μM, converted using a molecular weight of 309.45 g/mol. For naltrexone, total C max was based on 380 mg intramuscular injection of Vivitrol 1 every 4 weeks as the recommended dose (https://www.accessdata.fda.gov/drugsatfda_docs/label/2019/021897s045lbl.pdf). Total C max was 12.1 ng/mL or 0.035 μM when converted using a molecular weight of 341.41 g/mol. Given 21% plasma protein binding, free C max was 0.028 μM. For naloxone, total C max was the result of a 2 mg EVZIO 1 intramuscular/subcutaneous injection (https://www.accessdata.fda. gov/drugsatfda_docs/label/2016/209862lbl.pdf). Total C max was 7.91 ng/mL or 0.024 μM, converted using a molecular weight of 327.27 g/mol. In adult human plasma, 54% naloxone is in free, unbound form [27]. Thus, free C max of naloxone was 0.013 μM. Fig 1A shows a schematic diagram of a ventricular AP and contributions of the ionic currents examined in this study to different phases of the AP. In a highly simplified scheme, I NaP is responsible for AP upstroke (phase 0), I CaL and I NaL contribute to AP duration (APD; phase 2), and I hERG , thought to be the correlate of I Kr in cardiomyocytes, contributes to repolarization of the membrane potential (phase 3).  Representative current traces (upper panels) recorded from cells overexpressing the specified ion channel proteins and voltage protocols used to evoke them (lower panels). Control traces are shown in black; following bath application of a selective blocker, dark gray is used. For I CaL recordings in (C) and I NaL recordings in (D), I passive traces in light gray were calculated using input resistance values derived from the control traces (see "Methods"). I NaL and I NaP were studied using the same voltage protocol. I NaL was recorded in the presence of 150 nM ATX-II and its amplitude quantified as the peak inward current during the voltage ramp down phase. I NaP was recorded in the absence of ATX-II and quantified as the peak inward current during the -15 mV step. The selective blocker used to inhibit I hERG was 1 μM E-4031; I CaL , 100 μM verapamil; and I NaL and I NaP , 30 μM TTX. The arrowhead in each panel denotes the region where each current was measured to quantify drug effects.

Cardiac ion channel electrophysiology
https://doi.org/10.1371/journal.pone.0241362.g001 hERG1a/1b cells was faster than that in hERG1a cells (see summary spreadsheet for I hERG experiments at https://osf.io/tjuev/), consistent with different gating behavior for homomeric and heteromeric hERG channels [21]. Fig 1C shows current traces recorded from a representative cell expressing Ca V 1.2/β 2 /α 2 δ1 subunits, in control solution (black trace) and following bath application of the Ca V 1.2 channel blocker verapamil (100 μM; dark gray trace). The light gray trace is the calculated passive current (I passive ) for the control current trace. Despite mismatch at the +30 mV step, in the majority of the cells I passive and verapamil-insensitive current traces aligned well at the ramp-down voltage phase where the I CaL amplitude was quantified. Therefore, in this study I CaL amplitude was quantified by subtracting I passive . For a subset of cells, I passive and residual current following complete I CaL inhibition did not align well at the voltage ramp-down phase, suggesting the presence of an endogenous current. For these cells the residual current trace was used for current subtraction to estimate I CaL amplitude. Fig 1D and 1E show current traces recorded from representative cells overexpressing Na V 1.5α/β1 subunits, in the presence of 150 nM ATX-II to study I NaL (Fig 1D) and without to study I NaP (Fig 1E). Current trace recorded in control solution is shown in black; following 30 μM TTX application, in dark gray. In Fig 1D, I passive was calculated using the control trace and is shown in light gray. Note that the TTX-insensitive current trace closely matched I passive . Therefore, in this study I NaL was measured as the peak inward current at the ramp-down voltage phase and was quantified by subtracting I passive . I NaP was not I passive -subtracted given its large amplitude (in control solution: -7578.4 ± 865.4 pA; in TTX solution: -888.3 ± 148.6 pA; TTX-insensitive current: 10%; n = 25). Fig 2 shows the time course plots of representative pharmacology experiments for I hERG (Fig 2A), I CaL (Fig 2B), I NaL (Fig 2C), and I NaP (Fig 2D).
Multi-channel pharmacology for buprenorphine. Fig 3 shows the concentration-inhibition plots for buprenorphine against I hERG (Fig 3A), I CaL (Fig 3B), I NaL (Fig 3C), and I NaP ( Fig  3D). The IC 50 and n H values are summarized in Table 1. Buprenorphine's effects on homomeric and heteromeric hERG channels are not statistically different. Fig 3E shows the overlay of all concentration-inhibition plots and buprenorphine's free C max (0.00087 μM). There is no overlap between these in vitro and clinical concentrations. The ratios of IC 50 s for individual ionic currents and buprenorphine's free C max are presented in Table 1. This ratio for hERG1a, considered as the "safety margin" by drug developers, is 10,046. For reference, a safety margin of 30 to 45 and above has been proposed by several retrospective studies to indicate a low likelihood of drug-induced QT C prolongation [28][29][30].
Norbuprenorphine. Fig 4 shows the concentration-inhibition plots for norbuprenorphine against I hERG (Fig 4A), I CaL (Fig 4B), I NaL (Fig 4C), and I NaP (Fig 4D). Fig 4E shows the overlay of all concentration-inhibition plots and norbuprenorphine's total C max (0.22 μM). There is no overlap between these in vitro and clinical concentrations. Table 1 summarizes the values of IC 50 , n H , and the ratios of IC 50 s and total C max for norbuprenorphine. The safety margin of this drug, calculated using total and not free C max as the percent of protein binding has not been empirically determined, is 1,634.
Methadone. Racemic methadone, also known as R,S-methadone or (±)-methadone, was used in this study since this is the formulation approved in the United States to manage opioid addiction. Fig 5 shows the concentration-inhibition plots of methadone on I hERG (Fig 5A), I CaL (Fig 5B), I NaL (Fig 5C), and I NaP (Fig 5D). Fig 5E shows the overlay of all concentrationinhibition plots and methadone's free C max (0.44 μM). Unlike buprenorphine and norbuprenorphine, there is clear overlap between the in vitro and clinical concentrations for methadone. Table 1 summarizes the values of IC 50 , n H , and the ratios of IC 50 s and free C max for

PLOS ONE
methadone. The safety margin of methadone is 6.4. At free C max , methadone is estimated to inhibit I hERG by 16%, and I CaL and I NaL by 6%.
Naltrexone. Fig 6 shows the concentration-inhibition plots of naltrexone on I hERG ( Fig  6A), I CaL (Fig 6B), I NaL (Fig 6C), and I NaP (Fig 6D). The overlay of all concentration-inhibition plots and naltrexone's free C max (0.028 μM) shows no overlap between the in vitro and clinical concentrations (Fig 6E). Table 1 shows the values of IC 50 , n H , and the ratios of IC 50 s and free C max for naltrexone. The safety margin of this drug is 1,109.
Naloxone. Fig 7 shows the concentration-inhibition plots of naloxone on I hERG (Fig 7A), I CaL (Fig 7B), I NaL (Fig 7C), and I NaP (Fig 7D). Fig 7E shows the overlay of all concentrationinhibition plots and naloxone's free C max (0.013 μM). Table 1 shows the values of IC 50 , n H , and the ratios of IC 50 s and total C max for naloxone. The safety margin of this drug is 2,878.
iPSC-CMs. Drug effects on cardiac APs were tested using iPSC-CMs and MEA platform.      and naloxone (E) on MEA plates. The effect of opioids was analyzed 30 minutes after adding respective drugs to the wells. 'Quiescence' was observed in few instances at higher drug concentrations. These events are denoted as "Q" in panels along with the number of replicate wells that showed 'quiescence'. Arrythmia-like events were also observed in iPSC-CMs when treated with naltrexone at 10, 30 and 100 μM and naloxone at 100 μM. These events are denoted as "A" in panels along with the number of replicates. Data are represented as mean ± SEM. The vertical dashed line indicates free C max for buprenorphine, methadone, naltrexone, and naloxone, and total C max for norbuprenorphine.
https://doi.org/10.1371/journal.pone.0241362.g008 drug's free or total C max . Note that in these plots, the lines connecting the summary datapoints were for illustrative purposes and do not describe the relationship of how measured parameters change with increasing drug concentrations. At 1 and 3 μM, buprenorphine caused increases in ΔΔFPD in spontaneously beating iPSC-CMs. For reference, 1 μM is 1,111X higher than buprenorphine's free C max . Buprenorphine also caused reductions in ΔΔBP at 1 and 3 μM and ΔΔSA at 3 μM. At 10 μM and above, buprenorphine induced quiescence in all cells. Fig 8B shows the effects of norbuprenorphine. This drug also caused concentration-dependent increases in ΔΔFPD at 0.05-10 μM and ΔΔBP at 1-10 μM, and decreases in ΔΔSA at 0.05-10 μM. For reference, norbuprenorphine's total C max is 0.022 μM, and simulation predicted this metabolite to be extensively plasma protein-bound as the parent molecule (95 ± 5%; see "Methods"). Thus, the in vitro concentrations tested in these studies are likely much higher than the free drug level achieved with therapeutic dosing. Fig 8C shows the effects of methadone. The lowest concentration tested was 0.5 μM and approximately corresponds to free C max . Significant increase in ΔΔFPD is evident at this concentration. Methadone also caused concentration-dependent reductions in ΔΔBP and ΔΔSA, consistent with having effects on multiple cardiac ionic currents. These changes emerged at 1 μM (2.3X free C max ) and 0.5 μM (1.1X free C max ), respectively. Fig 8D and 8E show the effects of naltrexone and naloxone, respectively. Both drugs caused concentration-dependent increases in ΔΔFPD and ΔΔBP. For naltrexone, these electrophysiological changes emerged at 1 μM (28.6X total C max ); and for naloxone, at 3 μM (125X free C max ) for ΔΔFPD and 10 uM for ΔΔBP (416.7X free C max ). No effect on ΔΔSA was observed for either drug up to 10 μM. Naltrexone at 10, 30, and 100 μM and naloxone at 100 μM induced arrythmia-like events in 3-5 out of 8 wells and 2 out of 5 wells, respectively.

Discussion and conclusions
Buprenorphine and methadone are opioid receptor agonists that have been associated with QT C prolongation, with methadone generally accepted as having high TdP risk and buprenorphine having no documentation of associated TdP in the medical literature [1,26,[31][32][33][34]. This study shows that at clinically relevant concentrations, buprenorphine and its metabolite norbuprenorphine do not inhibit inward (I CaL , I NaL , and I NaP ) or outward (I hERG ) cardiac ionic currents (Fig 1) or alter AP characteristics in iPSC-CMs (Figs 3, 4, 8A and 8B). In contrast, methadone inhibits I hERG , I CaL , and I NaL , and increases FPD in iPSC-CMs (Figs 5 and 8C). The mechanism of QT C prolongation is thus opioid moiety-specific: for buprenorphine, this mechanism remains unresolved; for methadone, hERG channel block plays a role. Naltrexone and naloxone are two non-specific opioid receptor antagonists used to treat opioid use disorder and as emergency treatment for opioid overdose, respectively. Neither naltrexone nor naloxone has acute effects on the tested ionic currents or iPSC-CMs at clinically relevant concentrations (Figs 6, 7, 8D and 8E). In the BUTRANS 1 study, buprenorphine was not associated QT C prolongation when naltrexone was co-administered [1]. In light of the present results, this clinical observation suggests that QT C prolongation associated with buprenorphine is initiated by opioid receptor activation.
Buprenorphine has mixed effects on different opioid receptor subtypes. It is a partial agonist for MOR, an agonist for nociception or OR-like 1 (ORL1) receptor, and an antagonist at delta opioid receptor (DOR) and kappa opioid receptor (KOR) [35]. Table 2 shows the inhibition constants (K i ) or binding affinities of buprenorphine and other tested drugs on different human opioid receptor subtypes. In sheep Purkinje fibers and rat papillary muscles, buprenorphine at 1-10 μM reduced AP height and maximum rate of rise (V max ) while it prolonged APD [36]. These electrophysiological changes suggest inhibition of I NaP and I hERG , as demonstrated in subsequent patch clamp studies [10,11] (see Table 1 for IC 50 s). The present study confirms and extends the previous work, demonstrating that buprenorphine inhibits I hERG , I NaP , and additionally I CaL and I NaL , with IC 50 s in the micromolar to tens of millimolar concentration range (Fig 3A-3D; Table 1). Given the large difference between the in vitro concentrations and buprenorphine's free C max (Fig 3E), direct inhibition of these ionic currents seems unlikely following therapeutic dosing. Results from iPSC-CMs similarly show a lack of drug effect on repolarization at clinically relevant concentrations (Fig 8A). Together, these data demonstrate that the direct cellular targets of buprenorphine that mediate QT C prolongation are not hERG, Ca 2+ , or Na + channels or present on iPSC-CMs.
QT C prolongation by buprenorphine could also be mediated by its metabolite norbuprenorphine, which is a partial agonist for MOR and KOR, and a full agonist for DOR and ORL1 [37] (Table 2). Norbuprenorphine's IC 50 s against the tested ionic currents are in the micromolar to tens of micromolar concentration range, values that far exceed the drug's total C max following therapeutic dosing ( Fig 4E; Table 1). Experiments on iPSC-CMs show that at 0.005 μM (0.23X total C max ), norbuprenorphine caused no change in ΔΔFPD C , ΔΔBP, or ΔΔSA (Fig 8B). At 0.05 μM and above (>2.3X total C max ), norbuprenorphine caused concentration-dependent increases in ΔΔFPD C and ΔΔBP, and decrease in ΔΔSA. That ΔΔFPD C began to increase at 0.05 μM was unexpected based on the ion channel data presented in Fig 4. Nonetheless, this concentration is likely much higher than the free drug level after therapeutic dosing, as norbuprenorphine is predicted to exhibit extensive plasma protein binding as buprenorphine. Thus, effects of norbuprenorphine on cardiac ion channels or iPSC-CMs cannot explain QT C prolongation associated with buprenorphine. Naltrexone and its active metabolite 6β-naltrexol are competitive antagonists at MOR and KOR, and to a lesser extent at DOR [38] (Table 2). Naltrexone inhibited the tested ionic currents with IC 50 s in the tens of micromolar to millimolar concentration range (Fig 6A-6D), values far higher than the drug's free C max following therapeutic dosing (Fig 6E; Table 1). Results from iPSC-CMs similarly showed a lack of drug effect on APs at clinically relevant concentrations ( Fig 8D). These results suggest that naltrexone does not reverse or normalize QT C prolongation by buprenorphine by acting on hERG, Ca 2+ , or Na + channels or cellular targets intrinsic to iPSC-CMs. Rather, this drug likely prevents QT C prolongation by preventing buprenorphine from activating opioid receptors.
Naloxone is a competitive antagonist at MOR and KOR, and a weak antagonist at DOR [39] (Table 2). Direct electrophysiological effects of this drug on cardiomyocytes and cardiac ion channels have been reported in many studies. In the micromolar to sub-millimolar concentration range, naloxone has been shown to prolong ventricular APD [40][41][42][43][44] and decrease V max [41,42]-changes suggestive of drug effects on multiple cardiac ion channels. Indeed, in the tens of micromolar concentration range, naloxone has been found to decrease I Na , I to , I CaL , and I Kr , and increase I K1 (inward rectifier current that sets the resting membrane potential) in native myocytes [44], and inhibit I hERG in overexpression cells [45] ( Table 1). The present data are consistent with the literature, showing that naloxone inhibits I hERG , I CaL , I NaL , and I NaP with IC 50 s ranging from tens to hundreds of micromolar (Fig 6; Table 1). Given the drug's free C max of 0.013 μM, however, direct inhibition of the aforementioned cardiac ionic currents from past or present studies seems unlikely following therapeutic dosing. Consistent with this interpretation, naloxone at clinically relevant concentrations produced no change in ΔΔFPD C , ΔΔBP, or ΔΔSA in iPSC-CMs (Fig 8E).
Methadone is used as treatment of opioid use disorder and treatment for pain. While the literature states that this drug is a full agonist at MOR, following therapeutic dosing methadone can reach a level that binds to DOR and KOR as well ( Table 2). Methadone is associated with QT C prolongation and TdP [26], and was studied here to provide a comparison for buprenorphine. Past studies have found methadone to be a multi-ion channel blocker, with IC 50 s for I hERG [10,[46][47][48], I CaL [48], I NaL [49], and I NaP [48,49] all in the micromolar to tens of micromolar range. The present results are largely consistent with the literature (Fig 5; Table 1), showing that methadone inhibits I hERG by 16% and I CaL and I NaL by 6% each at the drug's free C max . The overall consequence of inward and outward current inhibition is increased ΔΔFPD C and ΔΔBP in iPSC-CMs at 0.5 μM (1.1X free C max ) (Fig 8C). Similar findings of repolarization delay by methadone have also been found in a study of sheep Purkinje fibers [50] and human stem cell-derived cardiomyocytes [48]. Thus, following therapeutic dosing, repolarization delay by methadone that results from acute hERG channel block can be expected.
An important limitation of this study is the lack of concentration analysis to confirm actual drug concentrations exposed to the recorded cells. Actual concentrations may differ from nominal drug concentrations due to issues related to stability, solubility, adsorption in the perfusion apparatus, and in the case of MEA experiments, the presence of serum proteins in the medium [51][52][53]. Such possibility has led to the question whether the central conclusion of this study-that buprenorphine and norbuprenorphine do not directly block hERG channels with therapeutic dosing-can still be supported. Two lines of reasoning suggest "yes". First, despite the uncertainty in in vitro drug concentrations, several retrospective studies have pointed to a hERG safety margin of 30 to 45 and above to identify drugs with low QT prolongation risk due to direct hERG channel block. Buprenorphine's safety margin is far greater than that threshold and excess of 10,000. Even norbuprenorphine's safety margin, determined using total rather than free C max , is excess of 1,600. Second, a previous study that compared drug loss for 39 compounds showed the most extensive loss to be 74% [51]. Even with that level of drug loss, buprenorphine's safety margin would still be 2600, whereas that for norbuprenorphine determined using total C max would still be 416. Therefore, the lack of concentration analysis in this study may affect the actual safety margin values for buprenorphine and norbuprenorphine. However, due to the magnitude of these values, the study conclusionthat acute and direct block of hERG channels by buprenorphine (and norbuprenorphine) is unlikely to occur with therapeutic dosing and cannot explain QT prolongation seen in Harris et al. (2017) [1]-is not altered.
Opioid receptor agonists have been shown to affect the release of neurotransmitters, including acetylcholine due to vagal stimulation [54,55] and noradrenaline due to sympathetic nerve stimulation [56,57] to the heart in basic science studies. Thus, it seems plausible that buprenorphine-associated QT C prolongation is mediated by activation of presynaptic opioid receptors that alter neuromodulatory tone to the heart, resulting in repolarization delay. In terms of opioid receptor localization in the human heart, one study has reported diffuse immunoreactivities for MOR, DOR, and KOR in ventricular myocardial cells [58]. These results seem to be at odds with the findings that buprenorphine does not alter electrical behavior of animal ventricular myocytes [36] or human iPSC-CMs (Fig 8A and 8B) at clinically relevant concentrations of nanomolar range. Additional studies with different opioid receptor antibodies to confirm receptor localization on human primary ventricular myocytes will be helpful to reconcile these discrepancies.
Sobanski and colleagues also reported strong immunolabeling for MOR, DOR, and KOR on nerve fibers in human ventricular tissues, with MOR and DOR immunoreactivities on sparse individual nerve fibers, and KOR immunoreactivity predominantly on intrinsic cardiac adrenergic (ICA) cell-like structures [58]. Double immunostaining showed that DOR colocalized with the sensory neuron marker calcitonin gene-related peptide (CGRP), a neural peptide that modulates cardiac and blood vessel functions [59]. Exactly which neurotransmitters (and co-transmitters) are released by nerve fibers that express opioid receptors, and if they translate to QT C prolongation, are important questions for future studies to address.
Another opioid with a large hERG safety margin and clinical QTc prolongation that crosses the 10 msec threshold of regulatory concern is hydrocodone (safety margin for I hERG > 1238). This drug is a full agonist at MOR with low affinity for DOR and KOR (https://www. accessdata.fda.gov/drugsatfda_docs/label/2019/206627s010lbl.pdf). The C max of hydrocodone following a 120 mg oral dose is 110 ng/mL or 0.24 μM, converted using a molecular weight of 449.4 g/mol. Based on 33% plasma protein binding, free C max of hydrocodone is 0.081 μM. The hERG channel study report for this drug's application package shows that at 100 μM, hydrocodone inhibited I hERG by 33% (https://www.accessdata.fda.gov/drugsatfda_docs/nda/ 2014/206627Orig1s000PharmR.pdf). The safety margin for hydrocodone is thus >1238. That there are opioids with mechanisms of QT C prolongation not involving acute hERG channel block has important implications regarding cardiac safety studies for this drug class. During early discovery, drug candidates are screened for hERG channel block to predict their likelihood of prolonging the QT C interval. A negative hERG signal for opioids thus may not be predictive of a lack of QT C prolongation. Pertinent to regulatory review of submitted cardiac safety data, there remain several outstanding questions for QT C prolongation observed with opioids. What is the extent of QT C prolongation by opioids in the absence of direct cardiac ion channel interaction, and does it saturate? Opioid receptors are downregulated in the continued presence of agonist. If this form of QT C prolongation indeed were mediated by opioid receptor activation, then does the QT C change track with receptor downregulation? How does the QT C response differ between first-time vs. chronic users of opioids? Regarding opioid use-associated TdP risk, postmarket experience for buprenorphine shows no association with TdP. In contrast, methadone, which directly blocks hERG channels, is associated with high TdP risk. Future studies that assess cardiac ion channel pharmacology, electrocardiographic effects, and postmarket data for additional opioids are important to evaluate whether the lack of association with TdP as seen with buprenorphine can be generalized to other drugs with QT C prolongation not mediated by acute hERG channel block.