Characterisation of mexiletine’s translational therapeutic index for suppression of ischaemia-induced ventricular fibrillation in the rat isolated heart

The ‘translational therapeutic index’ (TTI) is a drug’s ratio of nonclinical threshold dose (or concentration) for significant benefit versus threshold for adversity. In early nonclinical research, discovery and safety studies are normally undertaken separately. Our aim was to evaluate a novel integrated approach for generating a TTI for drugs intended for prevention of ischaemia-induced ventricular fibrillation (VF). We templated the current best available class 1b antiarrhythmic, mexiletine, using the rat Langendorff preparation. Mexiletine’s beneficial effects on the incidence of VF caused by 120 min regional ischaemia were contrasted with its concurrent adverse effects (on several variables) in the same hearts, to generate a TTI. Mexiletine 0.1 and 0.5 µM had no adverse effects, but did not reduce VF incidence. Mexiletine 1 µM reduced VF incidence to 0% but had adverse effects on atrioventricular conduction and ventricular repolarization. Separate studies undertaken using an intraventricular balloon revealed no detrimental effects of mexiletine (1 and 5 µM) on mechanical function, or any benefit against reperfusion-related dysfunction. Mexiletine’s TTI was found to be less than two, which accords with its clinical therapeutic index. Although non-cardiac adversity, identifiable from additional in vivo studies, may reduce the TTI further, it cannot increase it. Our experimental approach represents a useful early-stage integrated risk/benefit method that, when TTI is found to be low, would eliminate unsuitable class 1b drugs prior to next stage in vivo work, with mexiletine’s TTI defining the gold standard that would need to be bettered.

Animals and general experimental methods and materials. Rats were chosen for studies because they have a coronary anatomy that lends itself to generation of reproducible susceptibility to ischaemia-induced VF that is well characterised, and the in vitro (Langendorff perfusion) variation lends itself to the tractable evaluation of the concentration-dependence of the effects of drugs on VF and other variables 40 . Male Wistar rats (254-475 g) were anaesthetised with a lethal dose of sodium pentobarbitone (170 mg.kg −1 ) and heparin (160 IU.kg −1 ) via intraperitoneal (i.p.) injection to achieve a surgical level of anaesthesia, confirmed by removal of pedal and corneal reflexes. Hearts were then excised (causing death by exsanguination) and immediately submerged in cold (4 °C) Krebs solution containing NaCl 118.5 mM, CaCl 2 1.4 mM, glucose 11.1 mM, NaHCO 3 25.0 mM, MgSO 4 1.2 mM, NaH 2 PO 4 1.2 mM and KCl 3 mM. Hearts were then cannulated via the aorta for Langendorff perfusion via the coronary arteries, and perfused with Krebs (as above), filtered (5μm pore size), warmed to 37 °C and gassed with 95% O 2 and 5% CO 2 to achieve a pH of 7.4, at a constant pressure of approximately 80 mmHg. Mexiletine (Sigma-Aldrich, UK) stock solutions were made up in a vehicle of 1:4, Ethanol (VWR International, UK): Water (PURELAB ELGA Process Water, UK). Krebs perfusate salts were purchased from VWR International, UK. coronary artery ligation. A unipolar electrode was inserted in the apex of the heart, connected to a PowerLab (ADInstruments, UK), and used to detect heart rate and rhythm, with arrhythmias defined, detected and analysed according to guidance from The Lambeth Conventions 44 . A silk suture (4-0) was sewn around the left anterior descending (LAD) coronary artery, 1-2 mm below the left atrial appendage towards the apex, and threading both ends of the suture through a polythene tube. The suture was later tightened with and clamped (using forceps) to occlude the LAD coronary artery, inducing regional ischaemia, and subsequently loosened to reperfuse the artery. Following completion of the experimental protocol, the involved zone (IZ) size, defined as the anatomical region of myocardium that had been subjected to ischaemia, was identified by re-occlusion during perfusion with 5% disulphine blue dye (Patent blue VF sodium salt, Sigma-Aldrich, UK), dissected and quantified as % of total ventricular weight (TVW). An IZ size of 40-60% TVW was expected given the chosen suture positioning 45 . experimental design and protocol. Mexiletine concentrations were chosen to span values slightly below and above reported clinically effective free plasma concentrations for suppression of ventricular arrhythmias 46,47 .
Three concentrations of were examined: 0.1 μM, 0.5 μM and 1 μM (n = 18/group), all of which were compared with a matched vehicle control group (3 x n = 18). Initially, 1 µM was examined in the expectation this would suppress VF but possess ADRs. Once this was confirmed, the lower concentrations of mexiletine were studied in order to reveal at what concentration VF suppression could be achieved without ADRs. This experimental design was implemented to minimise animal usage. All experiments were blinded and randomised and followed a defined protocol.
Following a 5 minute period of Krebs perfusion, measurements of heart rate (beats.min −1 ), coronary flow (ml. min −1 .g −1 ), PR interval and QT 90 (QT interval at 90% repolarisation) interval times (msec) were begun, repeated at specified time points during the protocol. Perfusate was then switched to a test solution or control. After a further 5 minutes, regional ischaemia was induced for 120 minutes, followed by reperfusion for 10 minutes, and then IZ size delineation. In a subset of the hearts perfused with 1 μM mexiletine (n = 9) a higher K + concentration of 5 mM K + was introduced at the 90 th minute of ischaemia, with mexiletine or control solution continued for the remainder of the protocol, to examine whether any antiarrhythmic or adverse effects of mexiletine were dependent on ambient K + in the heart, a factor examined previously and found to influence outcome in studies with other class I (and class IV) antiarrhythmics [48][49][50] . The 120 minute length of ischaemia was also chosen to capture beyond the scope of the full spectrum of phase 1 ischaemia-induced ventricular arrhythmias in the rat isolated heart, which arise and then dissipate during the first 90 minutes of regional ischaemia 51,52 . The occurrence of VF during the first 90 minutes of ischaemia was presented and analysed separately from VF occurring later because there is a possibility that mechanisms of VF differ between early and more sustained ischaemia, and the actions of drugs diminished, exacerbated or even converted from protective to proarrhythmic 51,53 . Our protocol was designed to explore all these possibilities without adding unnecessary extra groups (minimising animal usage).

Coronary flow and ECG analysis.
Coronary flow (ml.min −1 .g −1 ) was measured by timed collection of effluent, which was weighed (1 g = 1 ml), and then expressed relative to the TVW of the respective heart. ECG and pressure measures such as heart rate (beats.min −1 ), PR and QT 90 intervals (msec), diastolic and developed pressures (mmHg) were obtained only during sinus rhythm. Ventricular arrhythmias defined according to the Lambeth Conventions II 44 were recorded as 'present' or 'absent' within sequential 30-minute time periods during ischaemia, and the during first 10 minute period of reperfusion. exclusion criteria. Any heart that, prior to switch to test solution, had a coronary flow that fell outside the range of 7 to 20 ml.min −1 .g −1 , a sinus heart rate of less than 200 beats.min −1 , or an IZ size less than 40% TVW was excluded from the study, and subsequently replaced to maintain equal group sizes while maintaining blinding 41 . Hearts that met exclusion criteria were not excluded if doing so would not affect the outcome of the study (i.e. if VF occurred despite the IZ being below the size criterion, which is used to ensure that if VF did not occur in an individual heart, it was not due to IZ being 'too small for VF to occur' 54 ).
Supplementary studies assessing left ventricular contractile function. Evaluation of left ventricular function was undertaken using Krebs modified to contain 4 mM KCl in an experiment undertaken separately from the arrhythmia study, owing to technical reasons recently identified 55 . A deflated intraventricular balloon (IVB), made from complaint non-elastic material, attached to a pressure transducer was inserted into the left ventricle via the left atrium for assessment of two components of contractile function, namely diastolic function as defined by diastolic pressure, and inotropy as defined by developed pressure (the difference between peak systolic pressure and minimum diastolic pressure during a single cardiac cycle during sinus rhythm). Once attached to the perfusion apparatus, hearts were submerged in coronary effluent contained within a heated jacket (37 °C) to preclude a fall in temperature, later, during global ischaemia. The IVB was filled with a small volume of saline (~0.02 ml) so that a pressure could just be measured, and the arbitrary value on the syringe noted as the 'zero volume' . Using the zero volume as a reference point the syringe was then filled with a volume of saline (typically ~0.12 ml for a 0.7-0.9 g heart) sufficient to generate a developed pressure >100 mmHg, without elevating diastolic pressure to >10 mmHg (avoiding excessive diastolic stretch). This volume was defined as the 'working volume' and was set when examining effects of solution switching and effects of ischaemia and reperfusion. After cessation of each protocol, the hearts were removed from the Langendorff set up, the atria and any extra-cardiac tissue was removed, and the TVW (g) was measured.
Experimental design and protocol. Mexiletine was evaluated at up to 5 μM on the expectation based on clinical reports that adverse effects on contractile function would require a higher concentration than that evoking adverse effects on the ECG 38 . Each group was compared with a vehicle control group (all n = 9). Following a 5 minute period of Krebs perfusion, measurements of heart rate (beats.min −1 ), coronary flow (ml.min −1 .g −1 ), ventricular diastolic pressure (mmHg) and developed pressure (mmHg) were taken at various predetermined time points. A Starling curve was constructed by deflating the IVB to zero volume before inflating the IVB with 0.02 ml increments of saline (up to a diastolic pressure <20 mmHg) and recording the incremental changes in diastolic and developed pressure (mmHg). Perfusate was then switched to one of the three test solutions and 5 minutes later a second Starling curve was constructed. Global ischaemia was then induced by clamping the tubing above the aortic cannula, maintained for 20 minutes. Reperfusion was initiated by releasing the clamp, using the same test solution, and maintained for 60 minutes, whence a third Starling curve was constructed to determine drug effects on recovery of mechanical function.
Data and statistical analysis. Selection of group sizes was based on previous experience with the models used 40 , and an expectation that assessment of threshold concentrations achieving effects (i.e., detection of weak drug effects) would require larger group sizes than normal for establishing statistical significance, as found in previous studies 41 . Studies were undertaken using randomization followed by blinded analysis 44 and in accordance with BJP's guidance in terms of group sizes, normalizations (none used) and Y axis labelling (all raw variables and no 'fold control') 43 . Statistical analysis was undertaken to determine whether individual test groups differed from drug-free controls. Gaussian distributed variables were subjected to an unpaired t test (when there were only two groups) or to ANOVA (1 way or 2 way, repeated measures when appropriate) when more than two groups were contrasted. After ANOVA, if F was significant and no variance inhomogeneity was detected, Dunnett's post hoc test was used to identify the groups in which variables had changed compared with controls. All Gaussian-distributed values were expressed as mean ± SEM since representation of raw data values or standard deviations renders many data sets impossible to fathom due to clutter. Binomially distributed variables were compared using Fisher's exact tests. The threshold P value denoting statistical significance was P < 0.05. Differences between groups are described as effects or changes always and only when statistically significant. Statistical analysis was performed in GraphPad prism 7 (UK) software. ethical approval. All applicable international, national, and/or institutional guidelines for the care and use of animals were followed.
Declaration of transparency and scientific rigour. This paper adheres to the principles for transparent reporting and scientific rigour of nonclinical research recommended by funding agencies, publishers and other organisations involved with supporting this research.

Results
Design validation. IZ size was similar in each group ( Fig. 1A-C), validating the effectiveness of randomisation in fulfilling the requirement for parity of arrhythmogenic stimulus 54 .

Concentration-dependence of effect of mexiletine on ischaemia-induced VF.
The incidence of VF was approximately 50% in controls, providing scope for detection of inhibition by mexiletine ( Fig. 2A-C). Mexiletine had no effect on VF at 0.1 and 0.5 µM whereas 1 µM abolished it, indicating a steep concentration-dependence ( Fig. 2A-C). Representative ECG traces of sinus rhythm, ischaemia-induced VT and ischaemia-induced VF are displayed in Fig. 2D.

ADRs.
Recorded variables in control hearts followed the pattern typical for the preparation 51,55-57 , with heart rate and coronary flow falling slightly during ischaemia, and then recovering during reperfusion, PR intervals varying little over time, and QT intervals increasing during ischaemia and decreasing during reperfusion (Fig. 3,  4 and 5A-D). Neither 0.1 nor 0.5 µM mexiletine had any effect on any variable (Fig. 3A-D). 1 µM mexiletine did not affect coronary flow (Fig. 5B), or heart rate versus control (Fig. 5A), but it did prolong the PR interval and QT interval with effects beginning before the onset of ischaemia (Fig. 5C,D), indicating off-target pharmacology in non-ischaemic supraventricular and ventricular tissue. The introduction of 5 mM K + at the 90th minute of ischaemia did not alter these effects.

Effects of mexiletine on left ventricular contractile function.
Mexiletine, tested only at the higher two concentrations (1 µM and 5 µM), had no effect on heart rate or coronary flow (Fig. 6A,B), replicating findings (above) from hearts not fitted with an IVB. Prior to ischaemia, neither concentration of mexiletine affected diastolic or systolic pressure at the IVB's working volume (Fig. 6C,D). In all groups, diastolic pressure was increased, and developed pressure was decreased during 60 min reperfusion following global ischaemia with the partial recovery of function during reperfusion not altered by 1 or 5 µM mexiletine (Fig. 6C,D). www.nature.com/scientificreports www.nature.com/scientificreports/ Consistent with these findings, Starling curve analysis revealed that neither 1 µM nor 5 µM mexiletine had any discernible effect on either systolic or diastolic function prior to ischaemia, or during reperfusion, with the slope of each volume:pressure line indistinguishable from control throughout the protocol (Fig. 7A-F).

Discussion
Mexiletine and the need for quantitative nonclinical therapeutic window data. Among the few drugs used to treat severe life-threatening ventricular arrhythmias during, or following, AMI 5,58 mexiletine is the archetypal orally-active class 1b antiarrhythmic 11 . It is the only clinically used antiarrhythmic for prevention of SCD that represents an archetype that may serve as a gold standard, since the only other drug for this indication, amiodarone, is anomalous within its nominal class 5 , and not an archetype that may serve as a focus for novel drug development. Clinically effective plasma mexiletine concentrations that provide antiarrhythmic benefit without treatment-terminating ADRs are reported to be 0.5-2.0 µg/ml 46,47 . To achieve this, an oral dose of 200-300 mg every 8 hours is required [16]. However, the plasma concentration range of a drug encountered therapeutically is often wider than the therapeutic window 13 . The average plasma concentration of mexiletine causing ADRs is 0.88 µg/ml, and the threshold concentration triggering treatment withdrawal is approximately 0.95 µg/ml 38 . The clinical therapeutic window is thus concerningly narrow. Unsurprisingly therefore, although initial clinical trials reported positive findings 31-35,37,39,59 mexiletine has not emerged as a drug for primary prevention of SCD 12 ; as with all antiarrhythmics its ADRs are perceived to outweigh benefit in this context 5,16 . Consequently, since this, the most useful drug has limited utility, there remains an important unmet clinical need for primary prevention.
In mexiletine's nonclinical development phase it would have been useful to have had a clear characterization of the doses and concentrations obtaining safe antiarrhythmic activity. Currently, for a drug to be viable for advancement to therapeutic evaluation in humans it must have a an adequate TTI 13 , and the drug with the best TTI would serve as a template with which to judge developing drugs. Unfortunately, although mexiletine was examined for antiarrhythmic activity in a variety of animal models in vitro 60 , ex vivo 61,62 and in vivo 27,[63][64][65][66] , no systematic attempt was made to determine its TTI. The aim of the present study was to rectify this, using the highly tractable rat Langendorff isolated perfused heart model and, in doing so, exemplify the model/approach and provide a mexiletine data set for use as a template with which to evaluate whether novel, as yet, untested drugs offer an advantage. www.nature.com/scientificreports www.nature.com/scientificreports/ Principal findings and implications. We found that 0.1 µM and 0.5 µM mexiletine did not exhibit antiarrhythmic effects. Unsurprisingly, these concentrations are lower than those (in plasma) reported to provide clinical effectiveness 46,47 . A doubling of concentration to 1 µM was effective against VF. However this concentration precipitated ADRs, causing prolongation of the PR and QT intervals, just as a doubling of plasma concentration  www.nature.com/scientificreports www.nature.com/scientificreports/ to 1 µg/ml results in serious ADRs in humans 38 , and subtherapeutic doses of mexiletine in a conscious rat model of ischaemia-induced VF caused an unacceptable reduction in blood pressure 67 . These results indicate that the consensus therapeutic index for mexiletine in humans (based on disparate data) was replicated by the TTI in  www.nature.com/scientificreports www.nature.com/scientificreports/ the rat Langendorff preparation, and was confirmed to be impracticably small, requiring a concentration above 0.5 µM yet below 1 µM -a TTI of less than two. This is clearly insufficient to meet the 30-fold criteria typically predictive of a safe, effective drug 14 .
Overall these findings validate the experimental approach elaborated in the present study and explain the limited clinical utility of mexiletine.
Study strengths and limitations. The Langendorff-perfused rat heart has been used extensively for evaluating drug-induced effects on ischaemia-induced VF, and is highly tractable 40 . Additionally, it provides scope for identifying drug-induced ADRs for numerous drug classes including those that block cardiac sodium current (I Na ) and cardiac and vascular L-type calcium current (I CaL ) 40,49,50,53,57,68 albeit not delayed rectifying potassium current blockers 40 . In addition to detection of ECG effects, increases in coronary flow in the perfused rat heart 50,69 are predictive of peripheral vasodilatation and clinical hypotensive effects 70 . The present approach therefore allows early identification of a new drug as unviable owing to dose limiting cardiac ADRs. However, an isolated perfused heart self-evidently cannot be used to assess noncardiac ADR risk.
Concern about possible noncardiac ADRs would arise if a drug were found to have a TTI in the rat isolated heart that is sufficiently large to justify further nonclinical development (i.e., >30), and noncardiac ADRs would then be examined by other means. The advantage of our approach, however, is that any drug with cardiac ADRs will have been excluded from further development. Thus, although mexiletine can cause central nervous system toxicity and gastrointestinal discomfort at plasma concentrations lower than those causing ADRs related to the ECG 31,34,35,[37][38][39]71 , its small TTI in the rat isolated heart renders its extracardiac ADRs irrelevant. For any new class 1 antiarrhythmic, a TTI of >30 would need to be met using our protocol, and only then would examination of its noncardiac ADRs be justified; reducing animal usage is an added benefit.
The rat heart has numerous advantages for TTI assessment of putative anti-VF drugs that are well documented, including an ability to manifest a reproducible susceptibility to ischaemia-induced VF 40 , but some hypothetical reasons for concern about this species choice remain. Class 1 drug-induced ventricular or AV nodal conduction abnormalities may occur to a greater extent, or at a lower drug concentration, in species such as the rat that has a basal heart rate higher than that of humans, if the drug's effects on I Na are characterized by slow binding and unbinding rates, typical of class 1c and 1a drugs 72 . However, this would not be a concern for drugs that bind and unbind from sodium channels quickly, such as mexiletine and other 1b drugs 72 . The present data accord with this since mexiletine's TTI in the rat heart mapped to the human therapeutic index. The predictive value of the rat heart may not be optimal for class 1c or 1a drugs, but these drugs are no longer used for treating Figure 7. Changes in Diastolic pressure (mmHg) and Developed pressure (mmHg) generated from Starling curves in the rat Langendorff preparation, with or without 1 µM mexiletine and 5 µM mexiletine performed during Krebs perfusion (A,D), after switch to 1 µM mexiletine, 5 µM mexiletine or vehicle (B,E) and following 20 min global ischaemia and 60 min reperfusion (C,F). The slope of diastolic or developed pressure (mmHg) and IVB volume (ml) was analysed by linear regression analysis. All n = 9 per group, mean ± SEM. There were no statistically significant differences between groups for any variable.
conclusions. Nonclinical data sets can be complex with a variety of estimates of affinity and potency for actions on a range of different molecular targets and tissues. Additionally, safety and effectiveness are rarely investigated in tandem. This makes it difficult to obtain an estimate of a TTI. For a decision to be made on the likely clinical success and safety of a new antiarrhythmic, the nonclinical database must be sufficiently accurate to mitigate against false inference. In the SCD therapeutic area, failed translation has become the rule rather than the exception, with development blighted by low effectiveness coupled with ADRs at therapeutic dosage [6][7][8][9][10] . This requires better early-stage elaboration of the TTI using simple tractable approaches.
The integrated evaluation of safety and effectiveness (exemplified in the present study) aims to reduce animal requirements by effectively halving the number needed to estimate the TTI if effectiveness and safety were evaluated independently. The rat Langendorff preparation was shown to detect cardiac ADRs contiguously with beneficial effects on ischaemia-induced VF with a TTI (<2) that replicates the consensus clinical therapeutic index. The experimental approach outlined was therefore validated and provides a template 'gold standard' for future early nonclinical assessment of novel antiarrhythmics, particularly antiarrhythmics with class 1b characteristics, the one remaining class that is not intrinsically flawed owing to insurmountable ADRs including proarrhythmia related to its primary molecular pharmacology in the heart.
Finally, it is important to clarify the context of the present work. If a novel drug (class 1b, certainly, or drug of novel class), were to be found to have a promising therapeutic index in the rat Langendorff preparation this would allow its development to proceed, and if the drug were later found to have dose-limiting noncardiac adversity this will be detected by other means, for example when the efficacy and safety are examined using an in vivo animal model. Species with a lower basal heart rate than the rat, such as the dog 73 are not suitable for early nonclinical evaluation owing to costs and ethical considerations. Our suggestion would be to utilise the Langendorff-perfused rat heart model to benchmark the TTI of a class 1b agent and verify findings in vivo, and then in a second species, only if justified at each stage of experimentation (TTI > 30-fold). Class 1b drugs that do have dose-limiting cardiac ADRs that give rise to a narrow TTI will be detected in the rat Langendorff preparation, and rejected, saving the animals, time and cost that would have been expended exploring the drug's ADRs by other means.