Re-assessing the risks of drug-induced arrhythmias during drug discovery

Can potentially lethal arrhythmias induced by novel drugs be better predicted by using more relevant and sophisticated electrophysiological screening methods? By Dr J Mark Treherne, Dr W Jack Scannell and Dr Richard C Saumarez.

Single-cell, electrophysiology-based screening systems are now routinely incorporated as pivotal decision-making assays, early into the drug discovery process to assess cardiovascular safety. Although such electrophysiological systems may be more predictive of potential cardiac safety and efficacy than non-functional binding assays, many apparently attractive new drugs are still being discontinued later in development. We suggest a new generation of screening assays, that better recapitulate the electrophysiology of the whole human heart, may allow better decisions across research and development. This article will review the challenges and explore the novel electrophysiological solutions required to better calibrate risk. We propose the evaluation of new electrophysiological recording and analysis techniques, using intact beating hearts and based on the fractionation analysis of paced electrocardiograms, to see if they have a higher degree of predictive validity in determining the risk of pharmacologically induced arrythmias in humans. Such techniques, if routinely deployed in drug discovery, may enable improved decision making in predicting serious adverse cardiovascular events. The objective of this proposal would be to identify the pro-arrhythmic side effects of potential new drugs for their earlier withdrawal from the lengthy drug discovery and development process, enabling the productivity of cardiovascular drug discovery to improve. Cardiovascular diseases are the leading cause of death globally.

Successful drug discovery is critically dependent on effective decision making during the lengthy research and development process that is required to discover new drugs. However, development pipelines across multiple therapeutic areas are still not meeting the anticipated clinical outcomes that are needed for bringing new and cost-effective drugs to patients. Back in 2012, it was observed that the number of new drugs approved per billion US dollars spent on research had halved roughly every nine years from 1950 to 2010, falling by around 80-fold in inflation-adjusted terms¹. The authors of this review introduced the concept of ‘Eroom’s Law’, which is Moore’s Law spelt backwards. The new ‘law’ referred to processes that are getting slower and more difficult to execute over time (the opposite of Moore’s Law, which in the 1960s modelled the rapid growth of the electronics industry). Unexpected late-stage failures in drug development are, of course, the costliest. A more recent analysis has illustrated how both assay validity and reproducibility correlate across a population of simulated screening and disease models². That review concluded that screening and disease models with higher “predictive validity” are a credible and practical solution to the problem. It argued that “perhaps there has also been too much enthusiasm for reductionist molecular models, which have insufficient predictive validity”, suggesting an increased role for more clinically relevant but perhaps, complex screening systems. Such assays could include, say, whole organs or animal models being incorporated as improved pivotal decision-making assays earlier into the drug discovery process. The overall conclusion of a more recent review³ proposes a range of options to improve predictive validity via the creation and, importantly, the rigorous evaluation of assay technologies.

This article develops these concepts with a focus on cardiovascular drug discovery and on cardiovascular safety pharmacology across multiple therapeutic areas. According to the World Health Organization, cardiovascular diseases remain as the leading cause of death globally, taking over 18 million lives each year. Specifically, we propose the hypothesis that whole-heart pacing studies are more likely than standard reductionist methods to recapitulate the biology of cardiac therapy (or toxicity) in human patients and hence have higher predictive validity than conventional methods. We also briefly explain how this hypothesis can be rigorously tested and how such pacing studies might then be incorporated into research and development processes.

The relevance of electrophysiology in drug discovery

Ion channels are implicated in many diseases and electrophysiological recordings are essential to understand their practical pharmacology⁴. Ion channels are also fundamental to understanding potential cardiovascular safety pharmacology issues for all drugs that are exposed to the heart. One of the most frequent adverse events, leading to numerous drug failures, are cardiac arrhythmias, which can lead to sudden cardiac death. For example, the discovery of the drug dofetilide⁵ has subsequently uncovered significant ion-channel- specific safety issues leading to sudden cardiac death, due to the drug’s adverse events observed in some patients⁶. Dofetilide was found to inhibit the human Ether-à- go-go-Related Gene (hERG) encoding cardiac potassium channel, which mediates the repolarising current in the cardiac action potential (IKr). Inhibition of hERG can lead to one of the most lethal arrhythmias, known as torsade de pointes. This term can be literally translated from the original French as a ‘twisting of peaks’ and is a specific type of abnormal heart rhythm that can lead to sudden cardiac death. It is a form of ventricular tachycardia (over 100 heat beats a minute) that exhibits those distinct characteristics on the electrocardiogram. These observations have led to the need for the early identification of hERG inhibition properties of biologically active compounds to screen out the downstream consequences. However, many other drugs exhibit a more complex multichannel pharmacology⁷, which is much harder to interpret and cannot be easily modelled by screening against a single ion channel.

While advances in laboratory automation have brought significant opportunities to increase screening throughput for electrophysiological cell-based assays targeting specific ion channels, a remaining cause of late-stage drug withdrawals is still pro-arrhythmia, even in non-cardiac related drugs that were never intended to target ion channels at all. This is a widespread problem, as approved drugs from many different therapeutic classes are known to cause or exacerbate a variety of arrhythmias⁸.

Furthermore, the extrapolation of pharmacological ion channel modulation to the clinical relevance of anti-arrhythmic drugs has also been problematic to interpret based on data only derived from simpler cell- based, often intracellular, electrophysiological screening assays. This article now proposes the use of some practical electrophysiological techniques recoding from intact and beating hearts as effective solutions to this longstanding problem. There is still a significant unmet need for new anti-arrhythmic drugs as well as for improved safety pharmacology.

Improving the predictive validity of complex electrophysiological assays

Significant improvement in the predictive validity of electrophysiological screening, assay systems are required, so that the data derived from such recordings can be translated from the effects seen in small mammalian hearts into the larger human heart. The pre-clinical pharmacological effects observed in such experimental model systems need to translate meaningfully into human clinical data to guide effective decision making in drug discovery. In particular, the pre-clinical assay systems should include screens that function as an intact organ, rather than just at a simpler cellular level, so that the spatial distribution of a potential arrhythmia substrate can be properly evaluated. The term ‘substrate’ in this context is defined as the conditions in heart muscle that can permit an arrhythmia to arise. Pathologically, arrhythmogenic substrates can be generated by a variety of pre-existing conditions that form a pre-requisite for the induction of an arrhythmia, which can only be studied effectively using more invasive electrophysiological techniques. Typical myocardial substrates include a combination of fibrosis, hypertrophy, ion channel dysfunction and abnormal calcium handling but the essential point is that the arrhythmia substrate depends on the spatial distribution of conduction and repolarisation abnormalities. The roles of individual mechanisms of electrophysiological fractionation analysis have also been studied in animal models in ferrets, rabbits and mice, which can be manipulated pharmacologically or genetically to illustrate relevant pathophysiological mechanisms.

Across drug discovery, cardiovascular safety pharmacology screening is a pivotal gating point for the progression of all drugs that are exposed to the heart. Dofetilide, for example, can be used to pharmacologically model the adverse effects seen in humans in isolated, perfused and paced ferret hearts⁹. Given that most lethal arrhythmias, including ventricular fibrillation and torsade de pointes, are re-entrant (as self-sustaining rhythm abnormalities) in humans, such assay systems need to examine the components of a re-entrant substrate. These components are conduction delay and temporary local block, both of which arise dynamically. Consequently, electrophysiological techniques need to be developed that can provide better translation from small mammalian hearts into larger hearts and then on into human clinical trials.

The potential benefits of paced fractionation analysis in drug discovery

Paced electrogram fractionation analysis or PEFA is an electrophysiological technique that has been initially developed for the evaluation human patients at most risk of sudden death. Further details are described in more detail below^9-13. The clinical applications of the technique in routine hospital settings have been recently reviewed¹³ and are illustrated diagrammatically in Figure 1. PEFA involves pacing the heart with a sequence containing increasingly premature extra stimuli and recording the response elsewhere in the heart. This technique was originally developed in patients with Hypertrophic Cardiomyopathy (HCM) and was based on the idea that the characteristic cellular disarray and fibrosis would cause multiple slowly conducting paths throughout the myocardium, as is illustrated in Figure 2. The refractory tissue would repolarise and then be invaded by the slowly conducting paths and create a re-entrant arrhythmia. As is shown in Figure 3, electrograms are recorded at increasingly short intervals between the stimuli. As the interval is decreased, the electrograms in high-risk patients become progressively longer and delayed with new potentials. The recorded electrograms were strongly associated with the risk of sudden death. This effect was also seen in other diseases (dilated cardiomyopathy, idiopathic ventricular fibrillation, post-myocardial infarct, the atria of patients with paroxysmal atrial fibrillation) and, in particular, the Long QT Syndrome (LQTS). LQTS is an inherited heart condition that in certain people causes fainting or seizures. Some people with LQTS do not exhibit any symptoms and may only become aware of their condition after having an electrocardiogram for another reason. In this syndrome, the myocardium is structurally normal and the majority of cases are due to abnormal gain of function of the fast inward sodium current and/or a reduction of outward potassium currents mediated by slowly and rapidly activating delayed rectifier potassium channels (known as Iks and/ Ikr, respectively). The mechanism of slowed conduction in the LQTS appeared to be due to local refractoriness, which below a critical coupling interval, disrupted normal activation and caused delay by activation finding its way around refractory tissue. Further investigation showed that there was a dose-dependent response of delayed conduction ex vivo in perfused small mammalian hearts with dofetilide.

In this study, a striking increase in fractionation with dofetilide was observed that was comparable to human clinical data⁹. The responses appeared to be equivalent to similar results obtained from patients with the specific LQT2 mutation, who had suffered ventricular fibrillation. The LQT2 subtype is the second-most common form of Romano–Ward syndrome (after the LQT1 mutation) and is responsible for 25 to 30% of all such cases. This is of considerable interest, since it suggests that drug-induced pro-arrhythmia can be investigated in isolated animal hearts as well as in humans using the PEFA techniques already developed to predict sudden death. In Figure 4, conduction curves from the perfused ferret heart in the absence and presence of 100nM dofetilide are shown in comparison with a conduction curve from a human LQT2 survivor and a control subject. Both conduction curves show clear similarities. Consequently, electrophysiological techniques have been developed that can provide the required translation from small mammalian hearts into larger mammalian hearts, including those in humans. Basically, the same PEFA technique can be adapted for those range of drug screening options from mice to men.

Next steps in evaluating the predictive validity of PEFA

Scannell et al.³ argued that the choice of screening and disease models often depends on tradition or availability. The models chosen can be surprisingly ‘under-evaluated’ given their powerful effect on the quality of research and development decisions. This, clearly, has major impacts on both the safety of patients and on the overall economics of research and development. The industry’s enthusiasm, or not, for screening and disease models often appears disconnected from serious empirical evidence on the ability of the models to predict drugs’ human-relevant performance³. Fortunately, there have been advances in the tools available to evaluate models^3,14,15 and in the rigour with which they are applied¹⁶. We now suggest applying such rigorous evaluation methods to PEFA.

The first step is to define the use of the model: the research and development decisions that it will inform. The next step is to then to draft a detailed specification of the features that a good model would possess. This can be thought of as a ‘Target Model Profile’^3,14,15 against which the PEFA model can be evaluated. Evaluation typically considers the extent to which the model recapitulates the relevant aspects of biology of human patients, how the tests and endpoints that can be measured using the model system map onto tests and endpoints that are relevant for human health. The ‘statistical and experimental hygiene’ of the model (typically, statistical power, signal-to-noise ratio, sources of bias) and the models likely ‘domains of validity’ also need to be evaluated. The domains of validity can be defined as the parameter space in which the model is likely to be predictive, and whether this is likely to extend to the intended human patient population. It should be possible to include in the evaluation a test of model performance against diverse set of drugs that are known to induce cardiac anomalies in human patients and, ideally, paired structural analogues that are known to be safe or that have little cardiac effect.

Depending on the research and development decisions that the PEFA models were intended to inform, one might cross calibrate small and large animal PEFA models against the same Target Model Profile, or examine the effects of animal age or co-morbidities on likely predictive validity.

Conclusions

By exploiting new emerging technologies in electrophysiology, such a PEFA, the earlier intervention of complex assay systems can potentially improve the rate of compound attrition in downstream development. PEFA-like techniques are now becoming ready to be deployed more effectively across the drug discovery process. Therefore, the pro-arrhythmic side effects of potential new drugs can be uncovered allowing for their earlier withdrawal from the lengthy drug discovery and development process. Furthermore, such techniques could also enable the discovery of novel anti- arrhythmic drugs to treat unmet medical needs.

Therefore, we propose that PEFA-based screening systems should be formally evaluated and, if they show high predictive validity, routinely incorporated as pivotal decision-making assays in drug discovery and development. Development pipelines across multiple therapeutic areas, including for cardiovascular diseases, can then better meet the expected clinical outcomes required for bringing safer and more efficacious drugs to patients.

However, a key question remains as to how new techniques, such as PEFA, can be introduced and practically implemented into existing drug discovery screening sequences or cascades to enable more effective decision making. Manual patch-clamping of single cells has become a mainstay of functional electrophysiology, often used to follow up after novel hit compounds are identified by non-functional fluorescence or ligand binding methods, which are considerably higher in their throughput. Various automated patch-clamp techniques have been introduced that are effective in the medium-throughput range for compound screening but generate similar data to more conventional manual systems. Overall, the electrophysiological screening of cell lines that express ion channels is typically used in the early stages of the drug discovery process to confirm the functional activity of novel compounds and elucidate their mechanism of ion channel modulation. The sequence, chronology and predictive validity of electrophysiological screening technologies and how they are applied in drug discovery is critical for success. Somewhat crudely, electrophysiology testing methods can be divided into either high-throughput but low-information assays or low-throughput but high-information assays. PEFA clearly fits in latter category and would be most effective later in the drug discovery process.

As a high-information assay system, PEFA is an antidote to placing too much pivotal decision making on reductionist cellular models. However, PEFA could be used to validate the analyses by machine learning algorithms and other artificial intelligence approaches of large datasets derived from more reductionist assays. Consequently, PEFA could play a significant role in development candidate prioritisation studies, which are typically evaluated in ex vivo organs and in vivo, whereby the predictive validity of detecting pharmacologically induced arrhythmias can be significantly improved. The use of ex vivo organs for screening drugs is not new, of course, but the use of PEFA is novel in this context of drug discovery.

The concept of ‘Eroom’s Law’, which was referred to above, is that drug discovery and development is getting slower and more difficult to execute over time and that this decline needs to be reversed. There are several potential explanations of this phenomenon that have been proposed¹. However, one such explanation seems to be the most appropriate for this article, has been described as the “brute-force bias”. This bias is the tendency to overestimate the ability of advances in brute force screening methods to show that a molecule is safe and effective in clinical trials. From the 1960s onwards, drug discovery has shifted away from classical pharmacological testing methods in, say, intact organs to reverse target-based approaches that result in the discovery of drugs that bind with ever higher affinities to the intended target protein. Despite being faster and cheaper in terms of unit cost, higher throughput approaches on their own may be less productive overall. The introduction of PEFA-like technologies may well help to redress that balance, thereby, enabling significant improvements in the productivity of cardiovascular drug discovery.

DDW Volume 24 – Issue 3, Summer 2023

References

1. Scannell et al. (2012) Nature Reviews Drug Discovery, 11: 191–200.
2. Scannell & Bosley. (2016). PLoS ONE 11(2): 1-21.
3. Scannell et al. (2022) Nature Reviews Drug Discovery, 21: 915–931.
4. Treherne (2006). Current Pharmaceutical Design, 12(4): 397- 406
5. Rassmussen et al. (1992). Journal of Cardiovascular Pharmacology, 20: S96-105
6. Aktas et al. (2007). Annals of Non- invasive Electrocardiology. 12(3): 197-202.
7. Johannesen et al. (2014). Clinical Pharmacology & Therapeutics. 96(5): 549-558.
8. Tisdale et al. (2020). Circulation.142: e214–e233
9. Saumarez & Grace. (2000). Cardiovascular Research 47: 11–22.
10. (2003). Journal of Physiology 552(2): 535-46.
11. (2002). Proceedings of the National Academy of Sciences (USA) 99(9): 6210-5.
12. (2006). Heart Rhythm 3(7): 771-8.
13. (2023).E PEuropace (in press).
14. (1992). Circulation 86(2): 467-74.
15. (2019). PLoSONE, 14(6): 1-17.
16. (2022). Communications Medicine, 2(1), Article 1: 1-16.

About the authors:

Dr Mark Treherne has been involved in the pharmaceutical industry for over 30 years and led a drug discovery research group at Pfizer’s research facility in the UK, including using electrophysiological techniques for screening compounds. He co-founded Cambridge Drug Discovery and then Xention, who automated electrophysiology assays for medium-throughput and high-throughput screens. Dr Treherne has a PhD in Pharmacology from the University of Cambridge. At the University of Basel, he developed recording techniques from the use in complex biological tissues in long-term culture. LinkedIn profile.

Dr Jack Scannell is CEO of Etheros Pharmaceuticals. He is best known for his work on R&D productivity, including the diagnosis of the causes of long-term negative productivity trends and, more recently, on the critical role of screening and disease model validity where “quality beats quantity.” He has extensive experience in drug and biotech investment (UBS and Sanford Bernstein), biotech drug discovery (e-Therapeutics), and consulting (BCG). He has a D.Phil. in physiology from Oxford University. LinkedIn profile.

Dr Richard C Saumarez is a fellow of the Royal College Physicians. He has a PhD in Electrical Engineering from the Engineering in Medicine Laboratory at Imperial College. He is the inventor of the PEFA technology described in this article and has demonstrated clinical proof of concept with multiple academic studies in 600 patients. He was a senior research associate and honorary consultant at Papworth Hospital in Cambridge and was also a member of the Engineering Faculty of the University of Cambridge. Dr Saumarez is the founding director of FEN EP. LinkedIn profile.