ILLUMINATING DRUG-INDUCED ARRHYTHMIA
MECHANISMS AND STEM CELL-DERIVED

CARDIOMYOCYTE HETEROGENEITY THROUGH
RAPID IONIC CURRENT PHENOTYPING

A Dissertation

Presented to the Faculty of the Graduate School

of Cornell University

in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

by

Alexander Phillip Clark

May 2023



© 2023 Alexander Phillip Clark

ALL RIGHTS RESERVED



ILLUMINATING DRUG-INDUCED ARRHYTHMIA MECHANISMS AND STEM

CELL-DERIVED CARDIOMYOCYTE HETEROGENEITY THROUGH RAPID IONIC

CURRENT PHENOTYPING

Alexander Phillip Clark, Ph.D.

Cornell University 2023

Cardiovascular disease is the leading cause of death in the United States, with

over 10% of these deaths attributed to cardiac arrhythmias. We believe novel ar-

rhythmia therapies developed over the coming decades have the potential to substan-

tially improve patient outcomes through the use of increasingly sophisticated precision

medicine tools. One such tool — induced pluripotent stem cell-derived cardiomy-

ocytes (iPSC-CM) — provides a laboratory model for patient-specific investigations.

These iPSC-CMs are currently used at academic medical centers to study arrhythmia

mechanisms, and by pharmaceutical companies to study the proarrhythmic potential

of drugs before clinical trials. While promising, these cells are also limited by their rel-

atively immature electrophysiological phenotype and cell-to-cell heterogeneity. Such

shortcomings have adversely affected the reproducibility, consistency, and depth of

insights from studies with these cells.

In this thesis, we develop a novel approach called rapid ionic current phenotyping

(RICP) that can be used to investigate the sources and extent of electrophysiological

heterogeneity during iPSC-CM patch-clamp experiments. We use this method to study

variations in physiology and patch clamp experimental artifact conditions that lead

to the heterogeneous phenotype of iPSC-CMs. Based on these findings, we propose

best practice methods to mitigate sources of electrophysiological heterogeneity during

experiments and address their knock-on effects during post-processing. We use the



RICP method in a drug cardiotoxicity screening pipeline, showing how we can acquire

surrogate markers of cardiotoxicity and identify proarrhythmia mechanisms from the

same iPSC-CM. Finally, we extend the RICP approach for use in an automated patch

clamp system to demonstrate its potential in a high throughput setup that is more

scalable in an industry setting. Ultimately, we believe RICP, together with insights it

has provided in this thesis, has the potential to affect basic science arrhythmia research,

and impact the way drugs are screened.



BIOGRAPHICAL SKETCH

Alex Clark graduated from the University of Virginia in 2014 with a Bachelor of Science

in Biomedical Engineering. He then attended Relay Graduate School of Education

where he received a Master of Education in Secondary Science. In 2018, after four

years in working in education, Alex enrolled in the Cornell Biomedical Engineering

Ph.D. program.

iii



ACKNOWLEDGEMENTS

I would like to thank my advisors, David Christini and Trine Krogh-Madsen for

their guidance and support over the last few years. From the beginning, they trusted

me as a scientist and treated me with respect as I pursued the problems I found most

interesting. Their mentorship has provided me with confidence when I present my

research to the broader scientific community and embark on a career that I hope will

lead to my own independent research program.

A special thank you to Siyu Wei. She has been in the lab since I joined in 2019. She

has taught me most of what I know about patch-clamp and has been an intellectual

and experimental partner throughout my PhD. This thesis would look very different

without her partnership and friendship. Thank you, Siyu!

I would like to thank my lab. Drew, Kristin, Radhika, Darshan, and Ivan have been

enormously supportive of my research, and provided me with plenty of advice along

the way.

I would like to thank my committee. Alessio and Peter have been encouraging

external advisors from the first time I presented my research to them. I am appreciative

of all their input, and their concern for me as a person first and scientist second.

I would like to acknowledge my family. My parents, Tom and Leisa, for nurturing

my love of science and providing me with everything I have needed to pursue my

intellectual interests. My older brother and sister, Brad and Hillary, who paved the

way for me, setting intellectual and athletic benchmarks that I strove, and still strive,

to match. My grandparents, Bryan and Carol, who modeled careers in biomedical

research that I dream to replicate.

Finally, I would like to thank my wife, Tracy. I had the great pleasure of coming

home to my brilliant and supportive wife at the end of each day. She forced me to time-

box my scientific endeavors so we could spend our evenings in dive bars, watching TV,

iv



and consuming chocolate chip cookies. Thanks for making life easy, buddy.

Ah, and last but not least, our faceless benefactor — the research reported in this

thesis would not have been possible without the support from the National Heart,

Lung, and Blood Institute of the National Institutes of Health under Award Number

F31HL154655 (to A.C.) and U01HL136297 (to D.J.C.).

v



TABLE OF CONTENTS

Biographical Sketch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi

1 Introduction 1

1.1 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Background 5

2.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.2 Cardiac electrophysiology and arrhythmias . . . . . . . . . . . . . . . . . 6

2.2.1 Cardiomyocyte electrophysiology . . . . . . . . . . . . . . . . . . 8

2.2.2 Cardiac arrhythmias and ion channel dysfunction . . . . . . . . . 11

2.3 The Comprehensive in vitro Proarrhythmia Assay . . . . . . . . . . . . . 13

2.4 iPSC-CMs as an experimental model . . . . . . . . . . . . . . . . . . . . . 15

2.4.1 iPSC-CM Electrophysiological Immaturity . . . . . . . . . . . . . 17

2.4.2 iPSC-CM ‘chamber-specific’ electrophysiology . . . . . . . . . . . 19

2.4.3 iPSC-CM heterogeneity . . . . . . . . . . . . . . . . . . . . . . . . 23

2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3 Methods and Approaches 27

3.1 Systems biology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3.2 iPSC-CM mathematical models . . . . . . . . . . . . . . . . . . . . . . . . 30

3.2.1 A brief history . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.2.2 Cardiac AP model formulations . . . . . . . . . . . . . . . . . . . 31

3.2.3 Simulations with cardiac AP models . . . . . . . . . . . . . . . . . 37

3.3 The patch-clamp technique . . . . . . . . . . . . . . . . . . . . . . . . . . 39

3.3.1 A brief history . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

vi



3.3.2 Patch clamp for cardiac electrophysiological recordings . . . . . . 40

3.4 IK1 dynamic clamp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

3.4.1 Automated patch-clamp . . . . . . . . . . . . . . . . . . . . . . . . 46

3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4 A novel drug screening pipeline identifies ionic proarrhythmia mechanisms 48

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

4.2.1 Pipeline design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

4.2.2 Voltage clamp protocol optimization . . . . . . . . . . . . . . . . . 53

4.2.3 Combining VC protocols . . . . . . . . . . . . . . . . . . . . . . . 54

4.2.4 iPSC-CM experiments . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.2.5 HEK-HCN1 Experiments . . . . . . . . . . . . . . . . . . . . . . . 57

4.2.6 Data analysis and statistics . . . . . . . . . . . . . . . . . . . . . . 59

4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

4.3.1 Optimizing a VC protocol to isolate individual currents for drug

cardiotoxicity screening . . . . . . . . . . . . . . . . . . . . . . . . 62

4.3.2 Synthetic maturation of iPSC-CMs by IK1 dynamic clamp im-

proves interpretability of iPSC-CM AP data. . . . . . . . . . . . . 62

4.3.3 IK1 dynamic clamp AP data identifies surrogate markers of car-

diotoxicity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

4.3.4 The optimized VC protocol qualitatively identifies drugs that

block greater than 30% of an ionic current. . . . . . . . . . . . . . 66

4.3.5 The optimized VC protocol identifies a previously unreported

quinine block of If. . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

4.4.1 Optimizing VC protocols for iPSC-CM drug experiments. . . . . 74

vii



4.4.2 Generating VC protocols that take advantage of the unique gat-

ing kinetics for each channel. . . . . . . . . . . . . . . . . . . . . . 75

4.4.3 Screening for drug cardiotoxicity using the novel VC protocol

and iPSC-CMs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

4.4.4 Limitations and future directions. . . . . . . . . . . . . . . . . . . 78

4.4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

5 Rapid ionic current phenotyping (RICP) identifies mechanistic underpin-

nings of iPSC-CM AP heterogeneity 81

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

5.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

5.2.1 iPSC-CMs are heterogeneous . . . . . . . . . . . . . . . . . . . . . 85

5.2.2 Rapid ionic current phenotyping provides insight into AP outliers 88

5.2.3 RICP identifies ICaL as driver of upstroke in depolarized cells . . 91

5.2.4 RICP identifies IKr-isolating segment as predictor of MDP . . . . 93

5.2.5 Seal-leak current likely contributes to the depolarized MDP . . . 96

5.2.6 RICP identifies strong outward currents as drivers of APD90 . . . 98

5.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

5.3.1 iPSC-CM heterogeneity confounds experimental results and lim-

its reproducibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

5.3.2 RICP is a tool to understand ionic current mechanisms of iPSC-

CM heterogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

5.3.3 ICaL likely drives upstroke in many iPSC-CM studies . . . . . . . 105

5.3.4 IKr is likely an important current in establishing MDP in depolar-

ized iPSC-CMs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

5.3.5 The optimized VC protocol elicits a large unidentified outward

current at 6 mV in some cells . . . . . . . . . . . . . . . . . . . . . 107

viii



5.3.6 Limitations and future directions . . . . . . . . . . . . . . . . . . . 108

5.3.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

5.4 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

5.4.1 iPSC-CM cell culture and electrophysiological setup . . . . . . . . 109

5.4.2 Calculating Iout during current-isolating segments . . . . . . . . . 109

5.4.3 iPSC-CM AP Models . . . . . . . . . . . . . . . . . . . . . . . . . . 110

5.4.4 Population of models and sensitivity analysis . . . . . . . . . . . 110

5.4.5 AP feature calculations . . . . . . . . . . . . . . . . . . . . . . . . . 110

5.4.6 Linear regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

5.4.7 Software and simulations . . . . . . . . . . . . . . . . . . . . . . . 111

6 Leak current, even with gigaohm seals, can cause misinterpretation of iPSC-

CM action potential recordings 112

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

6.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

6.2.1 Leak affects AP morphology even at seal resistances above 1 GΩ 115

6.2.2 Rseal is not stable . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

6.2.3 Rin is not a good approximation of Rseal at any holding potential . 120

6.2.4 Cm and Rin(0mV) correlate with minimum potential . . . . . . . 125

6.2.5 Fitting background currents in iPSC-CM models can absorb and

imitate Ileak . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

6.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

6.3.1 Leak affects AP morphology . . . . . . . . . . . . . . . . . . . . . 133

6.3.2 Predicting Rseal during experiments . . . . . . . . . . . . . . . . . 134

6.3.3 Correcting for Rseal during experiments . . . . . . . . . . . . . . . 135

6.3.4 Background currents absorb leak effects . . . . . . . . . . . . . . . 136

6.3.5 Modeling experimental artefacts . . . . . . . . . . . . . . . . . . . 137

ix



6.3.6 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

6.3.7 Limitations and future directions . . . . . . . . . . . . . . . . . . . 138

6.3.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

6.4 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

6.4.1 Modeled concentrations . . . . . . . . . . . . . . . . . . . . . . . . 140

6.4.2 Excluding Rs from Rin calculations . . . . . . . . . . . . . . . . . . 140

6.4.3 Genetic algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

6.4.4 Linear regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

6.4.5 Software and simulations . . . . . . . . . . . . . . . . . . . . . . . 142

6.4.6 iPSC-CM cell culture . . . . . . . . . . . . . . . . . . . . . . . . . . 142

6.4.7 Electrophysiological setup . . . . . . . . . . . . . . . . . . . . . . . 143

7 Rapid ionic current phenotyping demonstrates the potential and limitations

of automated patch experiments with iPSC-CMs 145

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

7.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

7.2.1 Experimental design . . . . . . . . . . . . . . . . . . . . . . . . . . 147

7.2.2 Developing an optimized VC protocol for RICP . . . . . . . . . . 149

7.2.3 iPSC-CM culture and dissociation . . . . . . . . . . . . . . . . . . 150

7.2.4 Patch clamp experiments . . . . . . . . . . . . . . . . . . . . . . . 151

7.2.5 IK1 dynamic clamp and other injected currents during AP record-

ings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

7.2.6 Linear regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

7.2.7 Software and simulations . . . . . . . . . . . . . . . . . . . . . . . 152

7.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

7.3.1 APs are short in duration and heterogeneous . . . . . . . . . . . . 152

7.3.2 Flecaininde and quinine reduce upstroke velocity . . . . . . . . . 154

x



7.3.3 INa block explains reduction in upstroke velocity . . . . . . . . . . 155

7.3.4 Lack of IKr in iPSC-CMs explains no change in APD90 . . . . . . . 158

7.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

7.4.1 APs are brief and heterogeneous in the automated patch setting . 161

7.4.2 Acquiring AP and rich VC data from the same cells . . . . . . . . 162

7.4.3 iPSC-CMs do not show evidence of IKr during automated patch

clamp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

7.4.4 Limitations and future directions . . . . . . . . . . . . . . . . . . . 163

7.4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

8 Conclusion and Future Directions 165

8.1 Summary of research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

8.2 Future directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

8.2.1 Cell specific modeling — a cautionary tale . . . . . . . . . . . . . 167

8.2.2 Effects of experimental artifact on iPSC-CM electrophysiology . . 170

8.2.3 The future of rapid ionic current phenotyping . . . . . . . . . . . 171

8.2.4 Limitations of automated patch systems . . . . . . . . . . . . . . . 172

8.2.5 We need to study iPSC-CM heterogeneity . . . . . . . . . . . . . . 173

A Appendix 175

xi



CHAPTER 1

INTRODUCTION

The heart consists of over 4 billion cells and contracts every second to pump

nutrient-rich blood to the body (Olivetti et al., 1995). The four chambers of the heart

must contract and relax in a specific sequence to ensure adequate blood flow to all

organs. Irregularities in the normal contraction-relaxation sequence are called arrhyth-

mias, and can impair the heart’s pump function. Progress has been made to reduce

arrhythmia mortalities over the last few decades, but they are still a leading cause of

death in the United States and significant morbidity burden.

Precision medicine will play an important role in improving arrhythmia outcomes

in the coming years as clinicians develop therapeutic strategies that consider increas-

ingly detailed genetic and physical characteristics of individuals.

Over the last decade, human induced pluripotent stem cell-derived cardiomyocytes

(iPSC-CMs) have become an essential tool in the study of patient-specific arrhythmia

mechanisms. These cells can be derived by drawing blood from an individual, induc-

ing cells from this blood into a stem cell state, and then replicating and differentiating

them into patient-specific heart cells, called cardiomyocytes. Prior to the discovery

of iPSC-CMs, testing on patient-specific cardiomyocytes could only be done through

a costly and invasive surgery that consisted of excising tissue from a patient’s heart.

iPSC-CMs, however, provide a renewable and patient-specific model for developing

and testing therapeutic approaches without the need for an expensive medical proce-

dure.

While iPSC-CMs offer great promise as a tool for studying cardiac arrhythmia

mechanisms, interpretation of experimental data from these cells is limited by their

1



immature and heterogeneous phenotype, and by the outsized role of experimental

artifacts during traditional electrophysiological studies. In this thesis, we study and

attempt to address issues that arise from these shortcomings. Specifically, we want to:

1. Understand the effects of experimental artifacts on iPSC-CM recordings so we

can design ways to compensate for them, and

2. Develop methods that provide mechanistic insight into heterogeneity, so this

property of iPSC-CMs can be taken into account during drug and genetic studies

With these goals in mind, we pursued four projects that build upon one another.

We first develop a novel experimental voltage clamp protocol that can isolate individ-

ual currents and can be used in iPSC-CM drug studies to determine proarrhythmia

risk and mechanism simultaneously. We then use this protocol to illuminate sources of

cell-to-cell heterogeneity caused by both experimental artifact and endogenous varia-

tions in ionic currents. We investigate the presence of an experimental artifact that is

ubiquitous in iPSC-CM single-cell electrophysiological studies and suggest remedies

for addressing its effects. Finally, we conclude by showing the potential and limitations

of extending the drug cardiotoxicity screening pipeline to an automated patch clamp

system that provides a high-throughput readout of drug effects on iPSC-CMs.

1.1 Thesis Outline

Below, we have provided an outline that walks you through the structure of this thesis.

In the next chapter, Chapter 2, we introduce arrhythmias, and discuss shortcomings

of the approaches used in iPSC-CM-based arrhythmia research.

2



In Chapter 3, we will discuss the experimental and modeling methods that we use

throughout the thesis. The chapter will provide brief histories of both cardiac ionic

current modeling and the patch-clamp technique.

In Chapter 4, we develop a novel voltage clamp protocol that can be used to isolate

individual currents in an iPSC-CM. We use this voltage clamp protocol as part of an in

vitro drug screening pipeline. We show how the voltage clamp protocol can be used

to provide a mechanistic explanation of drug cardiotoxicity, and identify novel ion

channel targets.

In Chapter 5, we use the voltage clamp protocol from Chapter 4 to study iPSC-CM

AP heterogeneity. The output of this protocol correlates with AP recordings, and pro-

vides an ionic current mechanism to explain a cell’s electrophysiological phenotype.

We call this method rapid ionic current phenotyping (RICP), and we use it to unravel

cellular mechanisms of AP generation.

In Chapter 6, we identify seal-leak current as a pervasive and unavoidable experi-

mental artifact present during single-cell iPSC-CM electrophysiological studies. Seal-

leak causes a depolarized and abbreviated AP, and varies from cell-to-cell, increasing

the presence of cellular heterogeneity and appearance of electrophysiological immatu-

rity. Based on these findings, we make recommendations that we believe can improve

the acquisition, analysis, and reporting of iPSC-CM AP data.

In Chapter 7, we extend this approach to an automated patch-clamp system to illus-

trate its potential in contexts (e.g., pharmaceutical industry) that require high through-

put readouts. We use AP and voltage clamp recordings to show how the electro-

physiological characteristics of cells in the automated patch-clamp environment differ

from manual patch-clamp recordings. We show the benefits and limitations of us-

3



ing the automated patch-clamp system for drug screening, with suggestions for the

types of drugs that can and cannot be screened using the system. We then discuss the

methodological advancements required before the approach can be used widely for

drug screening.

In Chapter 8, we discuss how each chapter contributes to the advancement of our

field and make suggestions for future research.

4



CHAPTER 2

BACKGROUND

Overview:
This chapter covers foundational physiological concepts relevant to the thesis, and

explores the use of human induced pluripotent stem cell-derived cardiomyocytes

(iPSC-CM) in arrhythmia research. The chapter begins with background on whole-

heart and single-cell cardiac electrophysiology and how perturbations to this system

can lead to arrhythmias. We then discuss how both genetics and drugs can affect cellu-

lar electrophysiological functions to increase the risk of developing a cardiac arrhyth-

mia. We end this chapter by discussing the limitations of iPSC-CMs as a model of

human physiology. This serves to motivate the research considered in the remaining

chapters.

2.1 Motivation

Cardiovascular disease is the leading cause of death and a significant contributor to

morbidity in the United States (Jiaquan et al., 2021). Over 10% of all cardiovascular-

related deaths can be attributed to cardiac arrhythmias. The prevalence of arrhyth-

mias, such as atrial fibrillation, are increasing in the United States as the population

ages. There is a need for tools and therapeutic strategies that can curb the burden of

arrhythmia-related morbidity and mortality in the United States. This public health

issue motivates the basic science research laid out in this thesis.

5



2.2 Cardiac electrophysiology and arrhythmias

The heart has two pumps, each consisting of two chambers: one atrium and one ven-

tricle. The right atrium receives de-oxygenated blood from the body and the right

ventricle pumps this blood to the lungs. The left atrium receives the oxygen-rich blood

that leaves the lungs and the left ventricle pumps the blood to the body.

The timing of atrial and ventricular contractions is highly coordinated to ensure the

ventricles: 1) contract in unison to apply the pressure required to move blood to the

lungs or body, and 2) relax to ensure adequate filling between contractions. Contractile

cells called cardiomyocytes make this contraction-relaxation cycle possible. Cardiomy-

ocytes, like neurons, are electrically excitable cells which means they are able to gener-

ate an action potential. An action potential occurs when an electrically excitable cell’s

membrane potential depolarizes, on its own or in response to a trigger, rapidly from

a negative potential (typically around -80 mV) to a positive potential (typically up to

40 mV). This depolarization leads to an influx of Ca2+ and increased Ca2+ intracellular

concentration, which enables the formation of crossbridges between actin and myosin,

giving rise to cardiomyocyte contraction.

Initiation of a heartbeat begins in the sinoatrial (SA) node (Figure 2.1). Cardiomy-

ocytes in this region beat spontaneously and play a critical role in setting the heart

rate. Depolarization of cells in the SA node starts an excitation wave that propagates

outward and causes the atria to contract. The excitation wave reaches the ventricles

through the atrioventricular (AV) node, which is responsible for slowing conduction,

allowing ventricles enough time to fill after atrial contraction. The signal then passes

to the left and right bundle branches, which split into purkinje fibers. The purkinje

fiber cells rapidly conduct the excitation wave throughout the inner layer of ventricu-

6



lar tissue, ensuring cardiomyocytes contract synchronously. Arrhythmias are caused

by disruptions to this complex and highly coordinated signalling system.

Figure 2.1: Action potential morphologies associated with regions of the heart.
Heartbeats initiate in the SA node, where cardiomyocytes beat spontaneously. This
leads to an excitation wave that causes atrial contraction. The excitation wave slows in
the AV node, and is then rapidly conducted through the bundle branches and purkinje
fibers. This rapid conduction ensures ventricular cardiomyocytes are excited in unison
to produce the pressure required to pump blood to the body.

7



2.2.1 Cardiomyocyte electrophysiology

The cardiac AP is the result of a complex interplay between ion-conducting proteins,

primarily classified as ion channels, pumps, or exchangers. While all three are essen-

tial for proper cardiac function, ion channels conduct the majority of current and are

responsible for most of the regional heterogeneity seen in APs across the heart (Figure

2.1). Ion channels are typically voltage-gated, with each channel having unique open-

ing and closing characteristics tied to the voltage across the cell membrane. Timing of

the opening and closing of these different ion channels affects the shape of an AP. The

cardiac AP has five phases, with each key ionic current playing a role in at least one of

these phases (Figure 2.2):

• Phase 0 – Rapid depolarization from a membrane potential near -80 mV to a pos-

itive potential is driven by sodium channels (INa) opening to allow Na+ to enter

the cell.

• Phase 1 – INa channels close as transient outward (Ito) potassium channels rapidly

open and close to cause a brief decrease in membrane potential.

• Phase 2 – Calcium (ICaL) and delayed rectifier potassium (IKr, IKs) channels open

and conduct opposing currents. This results in a small net outward current and

slow decrease in membrane potential. This is often referred to as the plateau

phase.

• Phase 3 – Calcium currents begin to fade as outward potassium currents remain

relatively large and drive rapid repolarization to a resting membrane potential

near -80 mV

• Phase 4 – In most cardiomyocytes, the inward rectifying potassium current (IK1)

conducts ions to maintain a resting membrane potential near -80 mV. SA nodal

8



cells have less IK1 and express hyperpolarization-activated and cyclic nucleotide-

gated (HCN) channels that conduct an inward current near -80 mV, slowly de-

polarizing the membrane potential and leading to the characteristic spontaneous

behavior of the cells.

Some of these currents are found in certain regions of the heart (e.g., If is not ex-

pressed much in ventricular cells). Such variations contribute to AP heterogeneity

that exists between different regions of the heart (Figure 2.1) and from cells within the

same cardiac tissue. For example, midmyocardial ventricular cells have a longer ac-

tion potential than either epi- or endocardial cells. This heterogeneity is an essential

part of cardiac physiology and required for proper functioning of the heart. It is even

protective against cardiac arrhythmias. However, small perturbations to this baseline

heterogeneity and underlying balance of ionic currents can be fatal. For example, dys-

function of IKr can lead to changes in AP shapes and increase the risk of lethal cardiac

arrhythmias.

9



−75

−50

−25

0

25

Vo
lta

ge
 (m

V)

−200

0

I N
a(

A/
F)

−5

0

I C
aL

(A
/F

)

0

5

I to
(A

/F
)

0

1

I K
r(A

/F
)

0.00

0.05

I K
s(

A/
F)

0 500 1000 1500
Time (ms)

0

1

I K
1(

A/
F)

0

1
2

3

4

Figure 2.2: Action potential and the major ionic currents contributing to its mor-
phology. Ionic currents are conducted through ion channels and pumps. Sodium- and
calcium-specific currents are inward (negative), and make the cell membrane poten-
tial more positive. Potassium currents are typically outward (positive), and make the
membrane potential more negative. A few ionic currents can be inward or outward
depending on the membrane potential and ion concentrations. This figure was gener-
ated using the Tomek et al. (2019) cardiac AP model.

10



2.2.2 Cardiac arrhythmias and ion channel dysfunction

Arrhythmias can be caused by abnormalities to numerous physiological processes.

Perturbations to molecular pathways (Balijepalli and Kamp, 2008), ion channel expres-

sion/function (Splawski et al., 2000), and heart structure (Prakosa et al., 2018) can all

be proarrhythmic. Many arrhythmias are multi-factorial, caused by abnormalities in

several physiological systems at once. For example, cocaine can lead to arrhythmias

by directly blocking sodium and potassium channels (Wood et al., 2009), and/or con-

duction issues caused by myocardial ischemia. Proarrhythmic abnormalities can be

acquired (as is the case with cocaine), congenital, or a combination of the two. Ac-

quired arrhythmias can be linked to chronic diseases such as heart failure (Selvaraj et

al., 2022), or acute changes to the system, such as alcohol consumption (Marcus et al.,

2021). Regardless of the predominant substrate, ionic currents always play a role in

contributing to or protecting against arrhythmias.

This thesis is concerned with the role these ionic currents play in arrhythmia for-

mation. While we often focus on drug-induced arrhythmias in this thesis, the methods

discussed in the chapters that follow can be used to study any arrhythmia that origi-

nates, at least in part, from ion channel dysfunction.

In this thesis, we use ion channel dysfunction to refer to changes in the amount or

opening/closing characteristics of a channel that results in physiological abnormali-

ties. These changes alter the normal balance of currents during the cardiac AP, and

subsequently, change the AP morphology in ways that can increase the risk of arrhyth-

mias. For example, long QT syndrome is a heart disease linked to a decrease in the

relative amount of repolarizing potassium currents (IKr, IKs) compared to depolarizing

calcium (ICaL) and sodium (INaL) currents (Amin et al., 2013). Long QT syndrome slows

the repolarization of cardiac tissue after contraction, which is a proarrhythmic sub-

11



strate linked to an increased risk of developing lethal cardiac arrhythmias, including

Torsade de Pointe (TdP).

The slowed repolarization seen in long QT syndrome patients can also be caused

by off-target drug effects. Terfenadine serves as a cautionary example. Terfenadine is

a non-drowsy antihistamine that was brought to market in the United States in 1985.

Over the following decade, it emerged that this seemingly innocuous drug used to

treat seasonal allergies increased the risk of sudden cardiac death in healthy patients

who were also prescribed ketoconazole, an antifungal (Monahan et al., 1990; Honig et

al., 1993). It was found that terfenadine blocks IKr at concentrations near its effective

free plasma concentration (Salata et al., 1995). The IKr block induces AP prolongation

and slows cardiac repolarization, directly leading to the increased risk of developing a

lethal cardiac arrhythmia. Importantly, this drug only slightly increased prevalence of

sudden cardiac death, which is why its proarrhythmic effects remained undetected for

over a decade. Terfenadine is one of 14 drugs that have been removed from the market

as a result of their potential to induce lethal arrhythmias (Stockbridge et al., 2013).

In response to the recall of these 14 drugs, in 2005, the international conference

on harmonization (ICH) designed and published drug safety guidelines that were

adopted by regulators and are used by pharmaceutical companies around the world

(U.S. Food and Drug Administration, 2005). The ICH S7B guidelines establish a set of

preclinical tests designed to identify drugs that cause lethal cardiac arrhythmias. Lab-

oratory tests outlined in ICH S7B focus on drug interactions with IKr, a channel that

is essential for proper cardiac repolarization, and is blocked by several of the recalled

drugs. These guidelines have been successful, effectively eliminating the approval of

drugs that induce lethal cardiac arrhythmias (Sager et al., 2014).

While highly sensitive, the ICH S7B guidelines are too conservative. The narrow

12



focus of these preclinical tests on IKr can result in false positive readouts for drugs that

target multiple channels. For example, verapamil is a safe and effective drug that has

been used for decades to treat several conditions, but is labeled proarrhythmic because

it blocks IKr. Verapamil also blocks depolarizing ICaL, which counterbalances the AP

prolonging effects of IKr block (Zhang et al., 1999). If verapamil were developed after

2005, there is a chance it would have been deemphasized or abandoned before clinical

trials.

Due to these shortcomings, regulatory agencies and pharmaceutical companies de-

termined there was a need for better preclinical drug screening assays that consider

other ion channels in addition to IKr. The goal of such assays are to: 1) improve the

specificity of drug screening without sacrificing sensitivity, and 2) improve mechanis-

tic understanding of a drug’s interaction with multiple cardiac ion channels before

approval.

2.3 The Comprehensive in vitro Proarrhythmia Assay

The Comprehensive in vitro Proarrhythmia Assay (CiPA) initiative was started by reg-

ulators and pharmaceutical companies, and its goal is to develop a new preclinical

drug proarrhythmia screening approach that addresses the shortcomings of the ICH

S7B guidelines. The initiative proposes the following three-step drug screening ap-

proach (Sager et al., 2014):

1. Conduct in vitro drug studies on seven cardiac ionic currents (IKr, IKs, Ito, INa, ICaL,

IK1, INaL).

2. Incorporate the ionic current blocking data from Step 1 into mathematical models

13



of cardiac action potentials to determine if the drug induces proarrhythmic AP

behavior.

3. Validate the findings using in vitro studies with iPSC-CMs.

The first step of CiPA requires the collection of dose-response data for each of the

seven ionic currents. This is done with several expression line cells, each transfected

with one of the ionic currents. Electrophysiological voltage clamp studies are then

conducted on the lines to see how each channel responds to progressively higher doses

of the drug. The percentage block of the channels is determined at each concentration.

The tested concentrations start close to zero and are increased to levels well above (e.g.,

>20x) the effective free plasma concentration. A recent study published dose response

data for the effects of 30 clinical drugs (Crumb et al., 2016) on each of the seven ionic

currents.

The second step is to incorporate these channel-specific effects into a mathematical

model of an adult cardiac AP (Dutta et al., 2017). Changes in the AP morphology (e.g.,

prolongation, slowing upstroke velocity) are used to predict whether a drug will be

proarrhythmic (Li et al., 2019).

The final step is intended to validate these computational findings by applying the

the drug to iPSC-CMs. Changes of the iPSC-CM AP morphology is compared to model

predictions, and the results from both are taken into account when judging the proar-

rhythmic potential of a drug (Blinova et al., 2018). Special attention is paid to discrep-

ancies between the in silico and in vitro results.

A key feature of the CiPA initiative is to provide mechanistic (e.g., block of 7 chan-

nels) insight to accompany arrhythmia predictions. This has become a focus of regula-

tory agencies, and is recommended for all new preclinical proarrhythmia assays (Li et

14



al., 2020).

This approach, and others inspired by it, provide accurate predictions of drug

proarrhythmia risk, with high levels of sensitivity and specificity (Tomek et al., 2019;

Li et al., 2019; Serrano et al., 2023). However, consistent and reliable iPSC-CM experi-

mental validation (Step 3) is imperfect (Blinova et al., 2018) due to several phenotypic

limitations of these cells.

2.4 iPSC-CMs as an experimental model

Human induced pluripotent stem cells (iPSCs) have emerged as a promising tool to

study human physiology outside of the clinic. iPSCs are developed by taking so-

matic skin or blood cells from a donor individual, and then reprogramming them to

a pluripotent state resembling cells from early embryos. These cells can be replicated

and directed to differentiate into various cell types, including cardiomyocytes, neu-

rons, and hepatocytes. The differentiated cells retain the genetic information of the

adult donor, providing a patient-specific model for laboratory testing. These cells are

used for affordable and non-invasive precision medicine studies that were not possible

before the development of iPSCs less than twenty years ago.

Cardiomyocytes derived from iPSCs have become one of the most widely available

and utilized differentiated cell types in both academia and industry. These cells are

used extensively to study cardiac development, congenital heart diseases, and drug

cardiotoxicity. They have been particularly useful to study proarrhythmic substrates

originating at the cellular level, including genetic mutations (Sala et al., 2019; Han et

al., 2014) and drugs that block cardiac ion channels (Liang et al., 2013).

15



While many studies have shown differences between control and experimental

groups, the depth of insights from iPSC-CM data is often limited by the cells’s elec-

trophysiological immaturity (Goversen et al., 2018b) and inter-/intralab heterogeneity

(Prajapati et al., 2021; Blinova et al., 2018). Cellular immaturity is present not only in

iPSC-CM electrophysiology, but also in the cells’s subcellular components and contrac-

tile machinery (Shinozawa et al., 2012). Single cell RNA sequencing has revealed vari-

ations in gene expression levels of iPSC-CMs from the same culture, providing some

explanation for the observed phenotypic heterogeneity (Schmid et al., 2020; Friedman

et al., 2018). Progress has been made to improve cellular maturity, but issues remain —

one hypothesis is that these cells lack the hemodynamic and structural inputs that are

present during normal cardiac development and likely required for maturation. The

most mature iPSC-CMs are now developed by simulating some of these inputs (Abu-

laiti et al., 2020), but a truly in vivo environment has not been replicated in a laboratory

setting.

This thesis is concerned with the ionic current underpinnings of electrophysiolog-

ical immaturity and heterogeneity in iPSC-CMs. Specifically, we want to better un-

derstand how both: 1) variations in ionic current densities, and 2) the presence of ex-

perimental artifacts contribute to the unpredictable electrophysiology that has come to

define these cells. Illuminating these mechanistic underpinnings has the potential to

improve our understanding of genetic- and drug-originating arrhythmias, and inform

cardiac differentiation and development strategies to improve cardiomyocyte matu-

rity.

We argue that the techniques developed in Chapters 4, 5, and 6 of this thesis should

be used during iPSC-CM experiments to evaluate the electrophysiological maturity

and heterogeneity of iPSC-CMs. Chapters 4 and 7 show how these strategies improve

16



the interpretation of iPSC-CM drug studies. While we apply these approaches in a

drug context, we believe they are well-suited for the study of genetic diseases and to

compare iPSC-CM differentiation protocols in studies of cardiac development. Addi-

tionally, we believe the findings gleaned from these novel approaches will be increas-

ingly informative as newer, more mature iPSC-CMs become available.

2.4.1 iPSC-CM Electrophysiological Immaturity

iPSC-CMs typically have an AP morphology resembling neonatal cardiomyocytes. De-

spite their considerable heterogeneity, iPSC-CMs have two AP features that almost al-

ways differ from adult cardiomyocyte behavior (Figure 2.3); they have 1) a depolarized

maximum diastolic potential (MDP) and 2) slower upstroke velocity (dV/dtmax).

0 100 200 300 400 500 600 700
Time (ms)

−80

−60

−40

−20

0

20

40

Vo
lta

ge
 (m

V)

Action Potential
Adult
iPSC-CM

−10 0 10
Time (ms)

Upstroke

Figure 2.3: Adult ventricular vs iPSC-CM AP. APs produced by mathematical model
simulations of an adult cardiomyocyte (Tomek et al., 2019) and iPSC-CM (Kernik et
al., 2019). The adult cardiomyocyte is hyperpolarized and has a much faster upstroke
velocity when compared to the iPSC-CM.

Adult ventricular cardiomyocytes come to rest below -85 mV and must be stim-

ulated to induce an AP. The depolarized MDP of iPSC-CMs is often attributed to a

17



relatively small IK1 (Figure 2.4). Recent studies, including our own (Clark et al., 2022a),

suggest that another contributor, seal-leak current, may also play a role in this depo-

larized MDP (Horváth et al., 2018; Van de Sande et al., 2021) — we will discuss and in-

vestigate seal-leak later in this chapter and in Chapter 6. The smaller upstroke velocity

is due to a lower sodium channel density, and decreased sodium channel availability

due to the incomplete recovery of channels at a depolarized MDP (Figure 2.4). While

upstroke velocity in adult cardiomyocytes is sensitive to changes in INa, in Chapter 5

we show that ICaL is likely the predominant current affecting AP upstroke velocity in

most of our iPSC-CMs. This is because the MDP of these cells are so depolarized that

nearly all INa channels remain inactive, while ICaL will conduct negative current even

at the depolarized potentials seen in many iPSC-CMs.

18



−75

−50

−25

0

25
Vo

lta
ge

 (m
V)

Adult

−75

−50

−25

0

25

iPSC-CM

−200

0

I N
a(A

/F
)

−200

0

−5

0

I C
aL

(A
/F

)

−5

0

0

5

I to
(A

/F
)

0

5

0

1

I K
r(A

/F
)

0

1

0.00

0.05

I K
s(A

/F
)

0.00

0.05

0 500 1000 1500
Time (ms)

0

1

I K
1(

A/
F)

0 500 1000 1500
Time (ms)

0

1

Figure 2.4: APs and current traces from adult (Tomek et al., 2019) and iPSC-CM
(Kernik et al., 2019) models. The current traces indicate that INa and IK1 are substan-
tially reduced in the iPSC-CM model. Such differences provide some explanation for
the depolarized MDP and slow upstroke in iPSC-CMs.

2.4.2 iPSC-CM ‘chamber-specific’ electrophysiology

Few studies have been more influential in shaping the community’s understanding of

iPSC-CM AP morphology and ion channel expression than Ma et al. (2011). It was the

first manuscript to report both iPSC-CM single-cell AP recordings and to characterize

19



major cardiac ionic currents.

It has served as a benchmark for iPSC-CM studies that followed and is the most

influential dataset in prominent iPSC-CM computational models (Paci et al., 2013;

Koivumäki et al., 2018; Kernik et al., 2019).

In Ma et al. (2011), they determined that their differentiation protocol produced a

mixed population of APs with three distinct phenotypes resembling atrial, nodal, or

ventricular cells (Figure 2.5). They group these cells using a shape factor that measures

AP triangulation:

Triangulation =
APD30 − APD40

APD70 − APD80

(2.1)

with ventricular-like >1.5 and nodal/atrial-like <1.5. Then, they separate nodal- and

atrial-like cells based on MDP (nodal > atrial) and dV/dtmax (nodal < atrial).

Figure 2.5: Atrial-, nodal-, and ventricular-like APs and summary statistics from Ma
et al. (2011).

This idea of a mixed phenotype cell culture with distinct chamber-specific car-

20



diomyocytes was not new at the time, as earlier iPSC-CM and embryonic stem cell-

derived cardiomyocyte studies similarly categorized cells based on AP morphology

(He et al., 2003; Zhang et al., 2009). While this categorization strategy provides a sense

of order, it also obfuscates the true heterogeneity of these cells and ignores the possi-

bility that iPSC-CM features may vary on a continuum rather than fit neatly into three

distinct groups.

The presence of three distinct groups within the Ma et al. (2011) dataset becomes

less apparent upon further investigation. To demonstrate this, we used the mean and

standard error values provided in their study (Figure 2.5) to create distributions of

dV/dtmax, MDP, and triangulation for atrial-, ventricular-, and nodal-like cells (Figure

2.6). The top row shows the individual distributions for each cell type and the bottom

row shows the combined distributions. When considering this combined distribution,

the appearance of three distinct groups is not obvious. These hypothetical distributions

demonstrate that the story of mixed phenotype iPSC-CM cultures is much messier

than is suggested by Ma et al. (2011) and others who followed (Ma et al., 2015; Garg

et al., 2019). The use of standard error of the mean and selection of three ideal cells

to display (Figure 2.5) is misleading and partly to blame for, what we believe to be,

the misconception that every iPSC-CM can be classified into a chamber-like group.

The story is likely more complicated. We believe it is prudent to operate under the

null hypothesis that iPSC-CMs are not chamber specific — such a framework is less

limiting, and allows us to consider phenotypes that may not fit neatly into any of these

categories.

21



−80 −60
0

2

4

6

8

10
Co

un
t

V
A
N

0 25 50 75
0

2

4

6

8

10

 
0 2 4

0

2

4

6

8

10

 
−80 −60

MDP (mV)

0
2
4
6
8

10
12
14

Co
un

t

0 25 50 75
dV/dt (V/s)

0

5

10

15

20

 

0 2 4
Triangulation

0

5

10

15

20

 

Figure 2.6: Hypothetical distributions of AP features based on (Ma et al., 2011)
dataset. MDP, dV/dtmax, and triangulation parameters were randomly selected using
the means, standard errors, and sample sizes for ventricular- (V), atrial- (A), and nodal-
like (N) cells from Ma et al. (2011). The top row shows the distribution of these param-
eters for each chamber-like grouping (e.g., V, A, and N). The bottom row shows the
distributions when the values are combined into one grouping. This illustrates that
there is no clear delineation between the three groups.

Protocols have been developed to direct iPSC differentiation towards ventricular

or atrial cell fates (Lee et al., 2017). These protocols are validated by measuring both

the electrophysiology and mRNA/protein expression levels. mRNA sequencing stud-

ies quantify the amount of atrial- (KCNJ3) and ventricular-specific (MYL2) genes in

a cell culture. Acquisition of AP morphology and current-specific IV data can then

be used as a phenotypic validation of mRNA/protein expression. For example, cells

from the atrial-directed protocol in Lee et al. (2017) have very short APD30 values and

22



show signs of IKACh, both of which are associated with the atrial cell type and agree

with their mRNA/protein data. While these protocols produce cells with more con-

sistent mRNA/protein expression profiles, and less variability in certain AP features,

heterogeneity in AP morphology remains (Feyen et al., 2020). To this point, we are not

aware of any protocols that result in a truly homogeneous cell population with highly

consistent AP morphologies.

This thesis is concerned with how to appropriately account for the effects of het-

erogeneity in studies of iPSC-CM electrophysiology, regardless of the differentiation

approach.

2.4.3 iPSC-CM heterogeneity

Figure 2.5 provides a nice example of the heterogeneity that often exists within a sin-

gle lab. This heterogeneity (e.g., MDP varying from -50 mV to -90 mV) is not unusual.

In a review of 24 datasets from 14 independent iPSC-CM studies, we found the MDP

standard deviation is >5 mV in 19/24 studies (Figure 2.7). Substantial heterogeneity

is observed in iPSC-CMs differentiated using protocols that produce mixed phenotype

populations and protocols that direct cells towards an atrial- or ventricular-like cell

fate (Lee et al., 2017). While variation in these parameters is smaller in iPSC-CMs dif-

ferentiated to optimize for cell maturity (Herron et al., 2016; Feyen et al., 2020), feature

variance remains large (e.g., MDP ranges of >10 mV).

23



100

200

300
C

2 4 6 8 10 12 14 16 18 20 22 24
0

20

40

60

dV
/d

t (
V/

s)

2 4 6 8 10 12 14 16 18 20 22 24

−80

−60

−40

−20
M

P 
(m

V)
A

2 4 6 8 10 12 14 16 18 20 22 24

200

400

600

800

AP
D

90
 (m

s)

B

1. Clark
2. Horváth
3. Es-Salah-Lamoureux (N)
4. Ma, D. (N)
5. Es-Salah-Lamoureux (A)
6. Giannetti
7. Lee (AI)
8. Ma, D. (A)
9. Lee (VI)
10. Han
11. Es-Salah-Lamoureux (V)
12. Ma, D. (V)
13. Ma, J. (N)
14. Garg (A)
15. Garg (V)
16. Doss (V)
17. Doss (A)
18. Van de Sande
19. Cordeiro
20. Herron (Immature)
21. Ma, J. (A)
22. Ma, J. (V)
23. Herron (Mature)
24. Feyen

Figure 2.7: Inter- and intralab AP feature heterogeneity. Mean and standard devia-
tions of action potential MP, APD90, and dV/dtmax from 24 independent datasets. These
datasets were taken from 14 different studies: (Clark et al., 2022b; Horváth et al., 2018;
Es-Salah-Lamoureux et al., 2016; Ma et al., 2015; Giannetti et al., 2021; Lee et al., 2017;
Han et al., 2014; Ma et al., 2011; Garg et al., 2019; Doss et al., 2012; Van de Sande et al.,
2021; Cordeiro et al., 2013; Herron et al., 2016; Feyen et al., 2020).

Heterogeneity across labs is also large, with mean MDP values ranging from -40 to

-80 mV. This is not surprising, as differentiation protocols differ from study to study —

some optimized for maturity, others designed for chamber specificity, etc. Interestingly,

of the studies that manually grouped cells into atrial, ventricular, and/or nodal (Ma et

24



al., 2011, 2015; Es-Salah-Lamoureux et al., 2016; Garg et al., 2019; Doss et al., 2012),

the lab of origin is more predictive of their MDP than the chamber categorization.

This further supports the idea that AP data from mixed-phenotype cell cultures should

not be grouped based on AP features. Even in the study that collects AP data from

ventricular- and atrial-specific cultures, the MDP for these datasets are more similar to

one another than to most other datasets (Lee et al., 2017).

Heterogeneity in iPSC-CMs has long been attributed to variations in the ionic cur-

rent densities. While it is true that channel densities differ from cell-to-cell, we posit

that variations in experimental artifacts also contribute to the inter- and intra-lab het-

erogeneity (Chapter 6).

The inconsistency in AP recordings acquired from iPSC-CMs within a single lab

and between labs is one of the greatest challenges of these cells and has limited their

utility in various contexts. For example, one study compared patient QT intervals

with the AP duration of iPSC-CMs derived from those same individuals (Blinova et

al., 2019). The variance in AP morphologies from each donor was so great that there

was no clear relationship between AP and ECG morphologies, and the data provided

no insight into patient-specific electrophysiology. In a different study, multiple labs

conducted the same drug cardiotoxicity experiments to study the reproducibility of

a drug screening approach. The results show multiple occasions where two labs re-

ported opposite responses to the same treatment (Blinova et al., 2018). Studies such as

these demonstrate the limited utility of iPSC-CMs caused by AP heterogeneity. And

too little is currently known about the differences in ionic current expression/function

that would give rise to such confounding results.

In Chapter 4, we develop a descriptive voltage clamp protocol that we then use

to rapidly phenotype individual cells in Chapter 5. The voltage clamp data can be

25



used to understand the ionic current variations present that cause intra- and interlab

heterogeneity and contribute to the limitations of these cells. The data can be used

to determine why two cells treated with the same drug produce opposite responses.

Ultimately, this rapid ionic current phenotyping can illuminate the sources of hetero-

geneity that limit the depth of insights during iPSC-CM studies.

2.5 Conclusion

Understanding the ionic current contributors to congenital and acquired arrhyth-

mias has far-reaching public health implications, but progress has been limited by a

dearth of laboratory models that recapitulate human physiology. iPSC-CMs offer great

promise as a tool for studying human, and more importantly, patient-specific arrhyth-

mia mechanisms. However, they are limited by their immature and heterogeneous

phenotype. Instead of obscuring this heterogeneity, in this thesis we try to study cell-

to-cell variances head on. Ultimately, we believe our work is a step towards deeper

understanding of these cells, with the potential to change the way people look at data

from iPSC-CMs. Such work has the potential to improve the interpretation of iPSC-CM

data, and therefore, the utility of these cells for studying drug-induced and congenital

arrhythmias.

26



CHAPTER 3

METHODS AND APPROACHES

Overview:
This chapter discusses the experimental and mathematical modeling techniques

used throughout the thesis. We start with a discussion of systems biology, and how we

use it to understand cardiac arrhythmia mechanisms. Then, we step through several of

the techniques used in this thesis, with brief historical discussions of cardiac modeling

and patch clamp. Understanding the history and details of these methods is helpful

to appreciate the significance of the methods developed as a part of this thesis and the

insights they can illuminate.

3.1 Systems biology

We take a systems approach to studying cardiac arrhythmia mechanisms.

The field of systems biology is often contrasted with traditional reductionist ap-

proaches to studying biological function. Historically, much of biology has focused on

developing a deep understanding of the component parts of a system — systems biol-

ogy, however, attempts to assemble these component parts into an ensemble that can

reproduce emergent biological behaviors.

Ron Germain, head of the Laboratory of Immune System Biology at the NIH, de-

fines systems biology as:

A scientific approach that combines the principles of engineering, mathematics,

physics, and computer science with extensive experimental data to develop a quantitative

27



as well as a deep conceptual understanding of biological phenomena, permitting predic-

tion and accurate simulation of complex (emergent) biological behaviors (Wanjek, 2022).

Systems biology enables the simultaneous study of many component parts. The

emergent behaviors of systems are often investigated by coupling computational sim-

ulations with in vitro experimentation, and together can provide insights into how a

system works. We find it helpful to use the 4Ms operational definition of systems

biology when we think about the relationship between computational and in vitro ex-

perimentation (Ideker et al., 2006), as shown in Figure 3.1:

• Model — This refers to the use of mathematical models that include compo-

nent biological parts and are built to recapitulate emergent biological behaviors.

These models are used to run simulations of biological systems, and the results

of which, serve as predictions to inform and prioritize experiments. In this thesis,

we use computational AP models (Kernik et al., 2019; Paci et al., 2018a; Tomek et

al., 2019) that have been built to recapitulate iPSC-CM or adult cardiomyocyte AP

behavior. In this chapter, we will discuss some of the component parts of these

models and the extensive experimentation that has gone into constructing them.

• Manipulate — Manipulating through experimentation provides insight into the

mechanisms of a system. We conduct in vitro experiments on single cardiomy-

ocytes and manipulate them using complex electrophysiological protocols and

channel-blocking drugs to study how the system will respond.

• Measure — Effects of a manipulation must be measured. With the patch clamp

studies used in this thesis, measurements are made nearly in real time, as cellu-

lar responses to complex electrophysiological protocols occur on the millisecond

timescale.

• Mine — The measured data is then analyzed and interpreted to provide insights

28



and make conclusions. Often, this mining results in findings that can be fed

back into improving the models. In Chapter 6, we show an example of just this.

Through mining of our patch-clamp data, we conclude that iPSC-CM models are

missing an experimental artifact equation, and so we add this to the model to

better recapitulate in vitro behavior.

Figure 3.1: The 4Ms of systems biology

In this chapter, we will step through the techniques and approaches we use to

model, manipulate, and measure the cardiomyocyte electrophysiological system. In

the next section, we will consider the modeling component of the 4Ms. We will dis-

cuss the history and mathematical underpinnings of the models we use in our lab and

throughout the thesis.

29



3.2 iPSC-CM mathematical models

3.2.1 A brief history

Much of modern electrophysiological modeling can be traced back to the seminal work

of Hodgkin and Huxley (1952), who published the first AP model (of the squid giant

axon). Through extensive electrophysiological experiments, they characterized the dy-

namic flow of sodium and potassium ions through the cell membrane, and linked these

flows to AP generation. Following their experiments, they constructed a mathematical

model with ordinary differential equations that described the dynamics of each ionic

current (i.e., the components of the system) that induced changes in the transmem-

brane potential throughout the AP (i.e., the emergent behavior).

Ten years after the work of Hodgkin and Huxley, Denis Noble published the first

cardiac ionic AP model (Noble, 1962). The model consisted of the same four ordinary

differential equations, but with parameter modifications that resulted in a morphology

that resembled APs recorded from specialized purkinje fibre cells from sheep and dog

hearts.

Over the years, cardiac AP models have grown in complexity as more ion-

conducting proteins have been discovered, each with its own multiyear story. For

example, in the decade that followed publication of the Noble (1962) model, multi-

ple groups identified calcium as playing a key role in the cardiac AP (Hauswirth et

al., 1969; Aronson and Cranefield, 1973; Reuter, 1967; Temte and Davis, 1967). These

findings exposed a missing component within the Noble (1962) study that was sub-

sequently addressed in a later model that incorporated an inward calcium current

(McAllister et al., 1975). This McAllister et al. (1975) model, constructed more than ten

30



years later, consists of ten ordinary differential equations and more than twice as many

parameters as the Noble (1962) model. There are similar stories in the mathematical

characterization of other ion-conducting proteins, such as funny current (DiFrancesco

and Noble, 1985), the sodium-potassium pump (DiFrancesco and Noble, 1985), and

two distinct delayed rectifier potassium currents (Zeng et al., 1995). The mathematical

formulations of these components relied on data acquired from traditional reduction-

ist biological experiments that isolated and extensively characterized these individual

currents.

Today, there are more than one hundred cardiac AP mathematical models, each

with its own benefits and limitations (Cherry and Fenton, 2008). There are different

cardiac AP models for guinea pigs (Luo and Rudy, 1994), dogs (Decker et al., 2009),

rabbits (Shannon et al., 2004), and many more animals. Such models have become

increasingly complex, with the CiPA human cardiac AP model including nearly 50

ordinary differential equations and >>100 parameters (Dutta et al., 2017).

In the next section, we will dig into the mathematical formulations of cardiac AP

components, and discuss how they are fit to recapitulate experimental data.

3.2.2 Cardiac AP model formulations

Acton potential models are typically constructed using the charge conservation equa-

tion (Hodgkin and Huxley, 1952):

Cm
dVm

dt
= −(Iion + Istim)

Here, Cm is the cell membrane capacitance, Vm is the membrane voltage, Iion is the

31



total transmembrane ionic current, and Istim is current from external stimulus. Iion is

a composite current consisting of individual currents that conduct K+, Na+, Ca2+, and

Cl- ions. In the Kernik et al. (2019) model, used throughout this thesis, there are 13

individual ionic currents that contribute to the net transmembrane current (Iion):

Iion = INa+ ICaL+ IKr + IKs+ IK1+ Ito+ If + ICaT + INCX + IPMCA+ INaK + IbCa+ IbNa

The flow of ions through each of these ion-conducting channels (e.g., rapid delayed

rectifier potassium, IKr), pumps (e.g., sodium-potassium pump, INaK), and exchangers

(e.g., sodium-calcium exchanger, INCX) can be mathematically modeled. The models

for most of these currents are formulated with a set of ordinary differential equations

that reproduce the opening and closing characteristics affecting ion flow. Often, new

cardiac AP models consist of current formulations inherited from previously published

AP models along with a few updates to bring the model in line with some experimen-

tal data (Niederer et al., 2009; Xu and Guevara, 1998). For example, the Kernik et al.

(2019) model consists of newly fit kinetic parameters for some currents (e.g., INa), but

unaltered parameters for others (e.g., INaK).

Formulations for ionic currents can vary in complexity depending on a current’s

unique opening/closing kinetics and the level of detail required to recapitulate exper-

imental data. For example, background sodium (IbNa) is modeled as a simple linear

current with no state variables and few parameters. ICaL, on the other hand, has more

complex dynamics and often requires >6 state variables with dozens of parameters.

The number of ions passing through a specific channel type in the cell membrane

typically depends on the density of the channel, the driving force for an ion, and the

probability that a channel is open. The general form for an ionic current formulation

32



is:

Ix = gxp(Vm − Ex)

Here, gx is the conductance or density of the channel (often in pS/pF), p is the

probability that the channel is open, and Ex is the Nernst potential for the ions passing

through the channel.

Most of the complexity of these models is in calculating the channel’s probability of

being open (p). Hodgkin and Huxley (Hodgkin and Huxley, 1952) style current models

are formulated with gating variables that can have values between 0 and 1. Currents

can have zero, one, or multiple gates. A Hodgkin-Huxley formulated INa often includes

three gating variables that we use in place of the p parameter:

INa = gNam
3hj(Vm − ENa)

The m parameter is raised to the third power to model the presence of three identi-

cal gates.

Ionic currents can also be formulated using Markov models. In contrast to

Hodgkin-Huxley (HH) formulations with independent gating probabilities (Figure

3.2), Markov models include discrete states that represent ion channel conformations

that either allow ions to pass through the channel (if open) or not (if closed or inac-

tivated). When run, the model keeps track of the proportion of ion channels in each

state. Using this information, it is possible to calculate the probability of a given chan-

nel to be open. This value replaces p in Ix = gxp(Vm − Ex). Markov models are often

more complex than Hodgkin-Huxley. The discrete states of the models can be used to

33



represent conformational details and drug interactions (Yang et al., 2020; Di Veroli et

al., 2013) that cannot be recapitulated with simpler HH formulations.

For both Hodgkin-Huxley and Markov models, the transition between states is cal-

culated using reaction rate coefficients. The rates of transition between an open and

closed gate in HH or states in a Markov model typically depend on the transmembrane

voltage and a set of kinetic parameters that are unique to the gate or state transition

(Figure 3.2).

34



Figure 3.2: Hodgkin-Huxley and Markov Models. Ionic current models are typically
formulated as Markov models or with Hodgkin-Huxley gating variables. Transition
between states in both types of models are calculated with reaction rate coefficients (k)
that depend on kinetic parameter values (a and b). Cardiac AP models are composed
of all the individual ionic current models. The change in transmembrane voltage over
time is equal to the sum of all ionic currents. Figure adapted from Whittaker et al.
(2020).

When cardiac AP models are built, the component ionic currents are fit first, and

then assembled into an AP model. The conductance parameters of each ionic current

are then fit and/or adjusted to reproduce AP features.

The first iPSC-CM model (Paci et al., 2013) was primarily fit to data from Ma et

al. (2011). While this model was built using the workflow described above, a few of

35



the ionic currents (e.g., sodium/calcium exchanger) were incorporated without adjust-

ment from a previously published guinea pig model Luo and Rudy (1994). This is not

unusual, as it is difficult to conduct all the electrophysiological experiments necessary

to fit every ionic current. While we hope these ”carry-through” currents are small con-

tributors to AP morphology, it is often unclear whether they are properly represented

in modern models.

Other iPSC-CM models have been developed (Koivumäki et al., 2018; Kernik et al.,

2019) and, in the case of Kernik et al. (2019), even include methods to reproduce in-

tercell variations in AP morphology that represent heterogeneity in iPSC-CM AP data.

This was an important advancement in the modeling of iPSC-CMs, and cardiac model-

ing more generally, as it explicitly recognizes the presence and importance of modeling

heterogeneity. Rather than providing a single model to run all simulations, they gen-

erate a population of possible cell models. We primarily use the average Kernik model

in this thesis, but believe the work of incorporating cellular heterogeneity into mathe-

matical models is important and should continue to be a focus of the field.

Limitations of the patch-clamp technique (see next section) make it challenging to

collect all the data necessary from a single cell to fit an iPSC-CM model. As such,

the iPSC-CM models discussed above include data from several datasets that, in total,

number in the hundreds of cells. The resultant models, therefore, represent an average

of data from many sources.

Given the heterogeneity in the Ma et al. (2011) dataset (Figure 2.5), it is clear these

models provide a reductive view of iPSC-CM electrophysiology. Even the Kernik et

al. (2019) model that includes a population of hypothetical individual parameter sets

does not attempt to validate that any one of these parameter sets reproduce behavior

from a real cell.

36



We believe there is a need to pursue patch-clamp methods that can rapidly phe-

notype the ionic current profile of a single cell, and see how this profile correlates to

AP features. This would provide information about a cell’s ionic current profile that

is explanatory of its AP morphology. Such a need motivates the development of our

voltage clamp protocol (Chapter 4) that we believe could be used to fit and validate

cell-specific models in the future.

As described below, while they have their limitations, cardiomyocyte models have

become an indispensable tool in the study of cardiac arrhythmias.

3.2.3 Simulations with cardiac AP models

Cardiomyocyte ionic current models are used to simulate behavior in many experi-

mental and clinical conditions. They provide an environment for generating and test-

ing hypotheses in a fraction of the time and expense compared to comparable in vitro

or in vivo studies. Computer simulations with mechanistic cardiomyocyte ionic current

models also provide the ability to track physiological processes and dynamics that are

not observable during experiments conducted in the real world.

Cardiomyocyte ionic current models are commonly used to study AP morphology

and the underlying ionic current dynamics responsible for generating the change in

voltage through time. Figure 3.3A shows an AP and calcium transient from the base-

line Kernik et al. (2019) model. While both the change in voltage and calcium transient

through time are observable during patch clamp studies, the computer model was run

in <1 s — an equivalent in vitro recording would require >1 hour. It is also possible to

simulate voltage clamp experiments (Figure 3.3B), where the transmembrane voltage

is set to various values through time, and ionic current is measured. During in vitro

37



experiments, it is only possible to directly measure the total transmembrane current

(Figure 3.3B, row 2). With models, however, it is possible to investigate all currents

that contribute to this composite Iion measure (Figure 3.3B, e.g., rows 3-4 show IKr and

ICaL).

−80

−60

−40

−20

0

20

Vo
lta

ge
 (m

V)

A

0 200 400 600 800 1000 1200
Time (ms)

0.0000

0.0002

0.0004

Ca
lci

um
 (m

M
)

−75

−50

−25

0

Vo
lta

ge
 (m

V)

B

−10

−5

0

I io
n(

A/
F)

−10

−5

0

5

I C
aL

(A
/F

)

800 1000 1200 1400 1600 1800
Time (ms)

0.0

0.2

0.4

I K
r(A

/F
)

Figure 3.3: AP and VC simulations from single-cell cardiac electrophysiology
model. A, Cardiac AP trace and calcium transient, which can be acquired experimen-
tally from iPSC-CMs using the patch-clamp technique and calcium sensitive dyes. B,
Voltage clamp traces from the same model. The voltage of the cell is initially clamped
to -80 and then stepped to 0 mV. During voltage clamp experiments, it is possible to
measure the total transmembrane current (row 2). This total transmembrane current
is a composite, consisting of the sum of all individual currents. Two of these currents
(ICaL and IKr) are displayed in rows 3 and 4. These individual currents cannot be mea-
sured directly during patch-clamp experiments.

These benefits of cardiomyocyte ionic current models allow for studies of arrhyth-

38



mia mechanisms that would not be possible otherwise. Single cell ionic cardiomyocyte

models have been used to provide mechanistic insights into arrhythmias originating

from inflammation (Campana et al., 2021), drug cardiotoxicity (Varshneya et al., 2021;

Cummins et al., 2014), genetic mutations (Kernik et al., 2020), and ion channel expres-

sion profiles (Varshneya et al., 2021). Additionally, single-cell cardiomyocyte models

can and have been incorporated into larger whole-heart electrophysiological models

and used to guide clinical decision making for patients who need cardiac ablation pro-

cedures (Prakosa et al., 2018; Boyle et al., 2019)

The continued refinement and validation of these models requires the use of in

vitro patch-clamp experiments. In the next section, we will discuss the patch-clamp

technique and cover the traditional types of current-voltage data collected and used to

construct key cardiac voltage-gated ion channel and cardiomyocyte models.

3.3 The patch-clamp technique

3.3.1 A brief history

The first measure of a nerve AP was recorded in 1868 by Julius Bernstein (Verkhratsky

et al., 2006). Bernstein estimated the resting membrane potential of the nerve to be

near -60 mV and theorized that this potential was maintained by selective permeability

of K+ ions. Following, Charles Ernst Overton proposed that Na+ was required for

the upstroke of the AP, and the AP was caused by an exchange of Na+ and K+ ions.

Overton also proposed the presence of a lipid membrane that separates the intra- and

extracellular spaces. This model was refined over the years — James Frederic Danielli

and Hugh Dawson proposed something close to our current model — a bilayer lipid

39



membrane with numerous pores that allow the passage of lipid insoluble ionic species

(Danielli and Davson, 1935).

Between 1930 and 1970, the giant squid axon was a popular model for studying

electrically excitable cells. Mini-electrodes were developed that could be inserted into

the axon and were used to measure the first recordings of the resting potential and

action potentials (Curtis and Cole, 1940; Hodgkin and Huxley, 1939). Within a decade,

the voltage clamp technique was developed (Marmont, 1949) — Hodgkin, Huxley, and

Bernhard Katz used the technique to describe a few of the major ionic currents required

for AP generation and the dependence of ionic fluxes on electrochemical gradients

(Hodgkin and Huxley, 1952; Hodgkin et al., 1952).

Over the two decades that followed, there was a desire to develop methods that

could measure the component ion channels giving rise to changes in voltage. In the

1970s, Ervin Neher and Bert Sakmann developed a method that used a micropipette

with tip surface area of < 10µm2 to measure the first single channel recording (Neher

and Sakmann, 1976). Less than a decade later, it was determined that a high resistance

seal could be achieved between the tip of these pipettes and the cell membrane surface

(Hamill et al., 1981). The formation of this seal provides the ability to rupture or perfo-

rate the cell membrane and to gain electrical access to the cell, while keeping the intra-

and extracellular spaces separate. This is the modern patch-clamp technique that we

use throughout this thesis to study cardiac electrophysiology.

3.3.2 Patch clamp for cardiac electrophysiological recordings

The traditional manual patch-clamp technique requires four components: the electri-

cally excitable cell (e.g., cardiomyocyte), an extracellular bath solution, a micropipette,

40



and an amplifier (Figure 3.4). The pipette is filled with an electrolyte solution, often

made to contain ionic species with concentrations similar to the intracellular space of

the cell. To patch a cell, the pipette tip is brought into contact with the external cell

membrane using a micromanipulator. Access to the cell can be gained by applying

negative pressure to rupture the cell membrane, or with a perforating drug that creates

channels for ions to pass through. This pipette contains an electrode and is connected

to an amplifier that can be used to manipulate and measure the electrical behavior of

the cell. An additional electrode is placed in the extracellular solution, completing the

circuit. The amplifier has two settings: it can be used in current clamp mode to mea-

sure the membrane voltage or in voltage clamp mode to measure the current passing

through the cell membrane (Figure 3.5).

Figure 3.4: Simplified patch clamp diagram. A, Two electrodes are connected to an
amplifier. One is in the extracellular bath. The other gains access to the inside of the
cell using the patch technique. B, A seal is formed between the tip of the micropipette
and cell membrane. An amplifier is used to manipulate and measure the electrical
behavior of a patched cell.

In voltage clamp mode, the amplifier clamps the membrane voltage to a user-

specified level and measures the ionic current passing through the cell membrane.

These voltage clamp protocols are specifically designed to probe the dynamics of indi-

41



vidual ionic currents. For example, a sodium-eliciting protocol will step from a large

negative potential (-100 mV) to a depolarized potential (e.g., -80 mV). This step from

-100 mV will be successively repeated at higher voltages (-75 mV, -70 mV,... 20 mV) to

study the voltage dependence of INa opening and closing. This is the data that is then

used to fit mathematical models (e.g., HH or Markov) of this ion channel. This type

of protocol represents the traditional reductionist approach to studying a cardiac ion

channel. This channel-specific investigation, however, has a couple of limitations.

First, these protocols often take a long time to execute. Because of this, it can be

difficult to acquire data about multiple ion channels from the same cell.

Second, individual ionic currents can overlap at various voltages, making it chal-

lenging to know the proportion of the current of interest. To address this issue, these

protocols are often used before and after the application of a channel-specific drug

(e.g., INa blocker). Assuming only the target channel is blocked, the difference between

the current traces before and after drug application is equal to the current of interest.

These protocols can take a long time to complete and channel-blocking drugs can

have detrimental effects on the cell after long exposure. Taken together, this prohibits

the study of many ionic currents from the same cell.

42



0 200 400 600 800
Time (ms)

−60

−40

−20

0

20

40
Vo

lta
ge

 (m
V)

A

−80

−60

−40

−20

0

Vo
lta

ge
 (m

V)

B

0 200 400 600 800 1000
Time (ms)

−3

−2

−1

0

1

Cu
rre

nt
 (A

/F
)

Figure 3.5: AP and voltage clamp data acquired from an iPSC-CM. A, Cardiac AP
trace recorded in current clamp mode using the patch-clamp technique. B, Current
response (bottom) of the same cell to a voltage command (top) stepping from -80 mV
to 6 mV.

Given the heterogeneity of iPSC-CMs, it is desirable to develop methods that pro-

vide insight into the presence and size of multiple ionic currents within the same cell.

Such protocols could be used to provide a mechanistic explanation for an individual

iPSC-CM’s AP morphology.

Sequential dissection is one such technique that has been used to study the pres-

ence and conductance of multiple ionic currents in the same cell (Banyasz et al., 2011).

The first step of sequential dissection is to acquire the baseline AP of a cardiomyocyte.

This AP is then used as the command voltage in voltage clamp mode. Current-specific

drugs are sequentially added to the cell and the baseline AP morphology is voltage

43



clamped after each application. Changes in the ionic current after the addition of each

drug provides insight into the presence and size of each ionic current. This protocol

has been used to study cell-specific behavior and compare cells within a heterogeneous

population. This approach, however, is limited in multiple ways: 1) because it requires

multiple drugs to determine the amount of each current, this approach cannot be used

to study the mechanism of action for an unknown drug; 2) the AP voltage clamp gen-

erates a current response with overlapping current contributions, especially during the

plateau phase; 3) the total sequential dissection protocol takes a long time to run, as the

addition of each drug requires a few minutes, and iPSC-CM patches only last for 15-30

minutes.

Given these limitations, there is a need for new protocols that can rapidly sample

the ionic current profile of iPSC-CMs without the use of channel-blocking drugs. This

need has motivated the development of voltage clamp protocols that take less than 30 s

to complete, and provide information on many ionic currents at once (Groenendaal et

al., 2015; Clark et al., 2022b; Lei et al., 2022). We will discuss this in Chapters 4 and

5. The data from such protocols can be used to provide a mechanism to explain cell-

specific AP morphology and drug cardiotoxicity.

3.4 IK1 dynamic clamp

The depolarized MP of iPSC-CMs has knock-on effects on multiple ionic currents that

play a role during the depolarization and repolarization phases of the iPSC-CM AP.

For example, INa recovers from inactivation at voltages below -60 mV, which is more

hyperpolarized than the MP for many iPSC-CMs. As such, INa is unable to adequately

recover in many cells, which results in a reduced availability of sodium channels, and

44



therefore a decreased dV/dtmax.

Historically, a depolarized maximum diastolic potential was dealt with by injecting

a constant hyperpolarizing current. This current could induce a resting membrane

potential below -65 mV providing for recovery of INa channels. However, because the

hyperpolarizing current is constant, it also affects phases 1 and 2 of the AP, leading

to a shortened duration (Verkerk et al., 2021). The shortening of APs caused by the

injection of this current can prevent the opening of important cardiac ionic currents,

and could hide the effects of drugs that typically induce AP prolongation.

IK1 dynamic clamp was designed to address the shortcomings of this constant cur-

rent approach. Dynamic clamp is a method characterized by the real-time evaluation

and injection of a simulated transmembrane current into an electrically excitable cell

(Wilders et al., 1996; Wilders, 2006). IK1 is largely responsible for establishing and main-

taining the quiescent resting membrane potential in adult cardiomyocytes, but is often

lacking in iPSC-CMs. IK1 dynamic clamp can be used to inject a simulated IK1 that

makes up for the lack of endogenous channels, and establish a resting membrane po-

tential below -65 mV. This dynamically clamped IK1 has kinetics that make it turn off

during most of the plateau and repolarization phases of the AP — as such, the AP will

not shorten as much as it does with the traditional constant current approach, and will

resemble a more electrically mature phenotype.

This is a strategy that we and others have used (Chapters 4, 7) to establish a resting

membrane potential below -65 mV in iPSC-CMs. These cells can then be paced (e.g., at

1 Hz) and AP morphology can be measured before and after various perturbations.

45



3.4.1 Automated patch-clamp

Automated patch clamp is a high throughput patch-clamp technique. In this thesis,

our automated patch-clamp data was acquired with a 4 channel Nanion Patchliner

(Becker et al., 2020).

The Patchliner system can be programmed to automatically execute most of the

manual steps that are required during traditional patch clamp experiments. To gain

electrical access to a cell, the Patchliner pipettes suspended cardiomyocytes into wells

with micro-pores that can suction the cell to form a seal and then rupture the cell mem-

brane. Once ruptured, preprogrammed commands execute to record voltage clamp

and AP data, including IK1 dynamically clamped APs. The Patchliner can also be pro-

grammed to apply drugs at multiple concentrations. This system benefits from the

use of a microfluidics system that allows for drugs applied directly to the cells — this

provides for more efficient washing on and off than is possible with traditional patch-

clamp setups.

In Chapter 7, we will show the data produced by this approach and discuss the

benefits (many of which are mentioned above), and drawbacks of using this automated

clamp system with iPSC-CMs. While these findings are promising, electrophysiologi-

cal differences of iPSC-CMs while in suspension limit the translational potential of this

approach in the near term. In the conclusion of this thesis, we will discuss the need to

understand the mechanism behind the differences observed in iPSC-CM data acquired

from manual vs automated patch-clamp setups. We hope that this motivates future

work on this topic.

46



3.5 Conclusion

Systems biology has a long history in the cardiac electrophysiology field. Mathemat-

ical models of cardiac APs have improved in unison with experimental techniques,

and together provide increasingly detailed insights into the cardiac electrophysiolog-

ical system. Today, computational models are even used in clinical decision making.

Building on a long history of arrhythmia research, in this thesis we develop methods

that we use to improve our understanding of single-cell electrophysiological hetero-

geneity and develop more realistic mathematical AP models.

47



CHAPTER 4

A NOVEL DRUG SCREENING PIPELINE IDENTIFIES IONIC

PROARRHYTHMIA MECHANISMS

This chapter is adapted from the following publication:

Clark, A. P., Wei, S., Kalola, D., Krogh-Madsen, T., and Christini, D. J. (2022).

An in silico-in vitro pipeline for drug cardiotoxicity screening identifies ionic pro-

arrhythmia mechanisms. British Journal of Pharmacology, 179(20), 4829–4843.

https://doi.org/10.1111/bph.15915

Contributions:

I led the design of the study, wrote the simulation codes, conducted the exper-

iments, and performed all of the analysis.

Overview:
This chapter encapsulates the majority of the work I completed during the first

three years of my PhD research. It also serves to motivate research in the remainder

of the thesis. Here, we show that by using IK1 dynamic clamp and rich voltage clamp

data, we can circumvent some of the issues introduced by iPSC-CM heterogeneity and

immaturity. However, these shortcomings do still limit the depth of insights that we

can draw from our data. We study and discuss these shortcomings further in the sub-

sequent chapters.

Here, we develop a novel computational method for optimizing a voltage clamp

protocol designed to isolate individual cardiac currents in iPSC-CMs without the use

of drugs. The voltage clamp protocol is less than 10 s, allowing us to apply it multiple

times throughout in vitro iPSC-CM studies. We acquire optimized voltage clamp and

48



AP data from iPSC-CMs before and after drug application and use them to identify sur-

rogate markers of cardiotoxicity and mechanism to explain changes in AP morphology.

This study represents a validation of the voltage clamp protocol and a successful real

world application (i.e., drug cardiotoxicity screening) of the RICP approach. We build

upon this validation of the RICP approach in the chapters that follow.

Before discussing the result, we will reintroduce many of the preclinical drug

screening and AP heterogeneity issues that were covered in Chapter 2.

4.1 Introduction

The use of in vitro IKr assays by regulatory agencies and pharmaceutical companies

prevents the approval of lethal proarrhythmic drugs (U.S. Food and Drug Adminis-

tration, 2005). However, the high sensitivity of such assays comes at the cost of low

specificity (De Bruin et al., 2005; Hancox et al., 2008; Gintant, 2011), and have been

shown to incorrectly label safe and effective therapies (e.g., verapamil and ranolazine)

as proarrhythmic (Johannesen et al., 2014; Crumb et al., 2016).

To address these specificity shortcomings, the Comprehensive in Vitro Proarrhyth-

mia Assay (CiPA) initiative was started in 2013 to guide the development of more

accurate preclinical tests (Sager et al., 2014). The initiative recommended a three-step

drug screening approach that includes: 1) quantifying drug effects on the major car-

diac ionic currents using expression system cells, 2) integrating these effects into in

silico models and using simulations to evaluate a drug’s proarrhythmic potential, and

3) validating simulations with human-derived induced pluripotent stem cell-derived

cardiomyocytes (iPSC-CMs) and human ECG studies.

Over the last decade, drug effects on iPSC-CM action potential (AP) duration or

49



field potential duration have become a common metric to evaluate drug risk (Navar-

rete et al., 2013; Blinova et al., 2018). iPSC-CMs however, are an imperfect model of

adult physiology, with an immature phenotype, high degree of heterogeneity, and de-

polarized maximum diastolic potential (Goversen et al., 2018b). These features make

it difficult to record consistent and reliable measures of proarrhythmia risk. Further-

more, the heterogeneity of these cells limits the ability to validate in silico simulations

and can produce discrepancies between experimental and computational results (Paci

et al., 2021).

Dynamic clamp of a synthetic inward rectifier potassium current (IK1) into iPSC-

CMs is a well-established method to improve the apparent electrophysiological matu-

rity of cells for drug studies. The IK1 model current is calculated in real time and in-

jected into iPSC-CMs to prevent spontaneous beating and establish a stable maximum

diastolic potential below -65mV (Quach et al., 2018; Fabbri et al., 2019). When paced

from this hyperpolarized resting membrane potential, cells have a more consistent,

and adult-like AP phenotype, making drug-induced AP changes easier to interpret

(Goversen et al., 2018a; Li et al., 2019).

In this study, we developed a pipeline that uses IK1 dynamic clamp and a novel

voltage-clamp (VC) optimization approach to determine both the proarrhythmia risk

and mechanism of a drug from iPSC-CM experiments. We start with in vitro iPSC-CM

studies to determine drug block that we then confirm with expression system dose-

response experiments. During the in vitro iPSC-CM studies conducted here, we ac-

quired optimized VC data, along with IK1 dynamically clamped APs, before and after

application of a CiPA-labeled low- (verapamil), intermediate- (cisapride), or high-risk

(quinidine) drug or quinine, which is also known to be proarrhythmic (Woosley et al.,

2021; Colatsky et al., 2016). In addition to correctly identifying all three ion channels

50



that were expected to be strongly blocked (>30%) by at least one of these drugs, the

protocol also identified a previously unreported block of funny current (If) by quinine.

This approach has potential as a new cardiotoxicity screening tool that can increase

specificity, while also providing information about the underlying mechanism.

4.2 Methods

4.2.1 Pipeline design.

Figure 4.1 displays the steps of a cardiotoxicity screening pipeline that we developed

and validated in this study. The first step in the pipeline is to use an in silico iPSC-CM

model-guided genetic algorithm (GA) to design a VC protocol that isolates individual

currents (Step 1). While the VC protocol could in principle be designed to isolate any

of the ionic currents present in the in silico model, in this study we focused on seven

currents that are most associated with AP morphology: IK1, If, slow (IKs) and delayed

(IKr) rectifier potassium, L-type calcium (ICaL), sodium (INa), and transient outward (Ito).

Optimized VC, as well as spontaneous and IK1 dynamic clamp and paced AP data, is

acquired from a patient-derived iPSC-CM before and after drug application (Step 2).

The IK1 dynamic clamp data is used to measure surrogate markers of cardiotoxicity

(Step 3), while the optimized VC data is used to identify ion channel targets (Step 4).

Dose-response data is then acquired for the identified targets using expression line cells

(Step 5). For example, in this study we acquired the dose-response data for quinine

block of HCN1, which further validated our findings on the unreported block of If by

quinine.

51



Time

Vo
lta

ge

No Drug
Drug

Time

Vo
lta

ge

IK1 DC

      Identify surrogate markers of  
cardiotoxicity (e.g. APD90 prolongation) 

%
 B

lo
ck

Concentration (μM)

High Dose

Baseline

      Develop dose-response curve 
from expression cell line experiments
5

3

2

Vo
lta

ge
Time

C
ur

re
nt

      Determine ion channel 
targets (e.g. quinine block of If) 
4

Acquire spontaneous, IK1 DC paced AP, and optimized VC data
Time

Vo
lta

ge

Kr CaL Na to K1 f Ks
1 Optimize a VC protocol with in 

silico models to isolate specific 
     currents (e.g. IKr, ICaL, ...)

Drug
Patient iPSC-CM

Figure 4.1: An in silico-in vitro pipeline to determine drug cardiotoxicity risk and
mechanism. Step 1, The Kernik-Clancy model with experimental artifacts is used to
develop a VC protocol that is specifically designed to isolate currents. Step 2, Spon-
taneous, IK1 dynamic clamp and paced AP, and optimized VC data is acquired from a
patient-derived iPSC-CM before and after drug application. Step 3, The change in IK1

dynamic clamp and paced AP data from pre- to post-drug application is used to iden-
tify AP prolongation, a surrogate marker of cardiotoxicity. Step 4, Changes in VC data
is used to determine the ion channels targeted by a drug. Step 5, After identifying the
ion channel targeted by a drug, a dose-response curve is developed for each of these
ion channels using expression line cells.

52



4.2.2 Voltage clamp protocol optimization

The baseline Kernik-Clancy iPSC-CM model was used in this study (Kernik et al.,

2019). Experimental artifacts (e.g., seal leak, series resistance compensation, and volt-

age offset) were included in our model simulations following the simplified artifact

model from Lei et al. (2020a). Taking experimental artifacts into account produce bet-

ter fits of ion channel models to experimental data, while adjusting fewer parameters

(Lei et al., 2020a). The effects of these patch-clamp artifacts are particularly pronounced

within the first few milliseconds after a voltage step (Figure A.1). Prior to the VC proto-

col optimization, the model was run to steady state, and then simulated under voltage

clamp at -80 mV for 20 s.

An optimized VC protocol was designed for each of seven currents (IKr, ICaL, INa,

Ito, IK1, If, and IKs). Custom Python code that included the DEAP Python package was

developed to implement the VC optimization GA (Fortin et al., 2012). The GA had 200

individual protocols per generation and 50 generations. Each protocol in the GA had

a set of four voltage segments. Each segment could be either a step or ramp between 5

and 1000 ms long and was constrained to voltages between -120 and 60 mV.

To evaluate the fitness of a VC protocol, the Kernik-Clancy model with experimen-

tal artifacts was clamped, and the percent contribution (C(t)) of the target current was

calculated at every timepoint.

C(t) =
|Ix(t)|

Σn|In(t)|

In this equation, Ix is the target current. The denominator is the sum of the absolute

values for all currents, including ionic currents and pipette leak (from the artifact equa-

tions). The best possible contribution score is equal to one and represents when the

53



target current contributes all the current observed during a patch-clamp experiment.

We calculated the average contribution, C(t), over a 10 ms window at each timepoint

and used the highest contribution window value as the fitness score for the protocols.

4.2.3 Combining VC protocols

The protocol with the highest current contribution for each of the seven ionic currents

(Figures A.2-A.8) was combined into one large protocol. Before combining the proto-

cols, they were systematically shortened using a two-step process. First, the portion

of the protocol >50 ms after the maximum current contribution window, identified by

the fitness function, was removed. Then, 10 ms segments were incrementally removed

from the beginning of the protocol, while ensuring the max current contribution did

not decrease by more than 5%. The seven shortened protocols were connected by 500

ms holding steps at -80 mV. The 500 ms duration was chosen for two reasons: 1) it is

long enough for many ion channel state variables to reach steady state and 2) the max-

imum contribution of each current did not decrease by more than 10% for any of the

currents (Table A.1).

To validate that the VC protocol isolates current during the same time windows in a

cell with different conductance and kinetic parameters, the optimized VC protocol was

applied to the Paci iPSC-CM model (Paci et al., 2018b) with experimental artifacts. The

VC protocol was also applied to the Kernik-Clancy model with extracellular Ca2+ con-

centration set to 1.2 mM, a value within the normal physiological range, but different

from our extracellular patch-clamp solution. We found little difference in the amount

of current isolation under these two conditions (Table A.2).

54



4.2.4 iPSC-CM experiments

Frozen vials of iPSC-CMs were thawed and cultured as a monolayer in one well

of a 6-well plate precoated with 1% Matrigel and supplemented with RPMI media

(Fisher/Corning 10-040-CM) containing 5% FBS (Gibco 16000069) and 2% B27 (Gibco

A1895601). Cells were placed in an incubator at 37°C, 5% CO2, and 85% humidity for

48 hours. When replating, cells were lifted with 1 mL Accutase (Corning A6964), and

the enzymatic reaction was blocked with DMEM/F12 (Gibco 10565-042) plus 5% FBS

(Burridge et al., 2014). Cells were diluted to 100,000 cells/mL and re-distributed to 124

sterile 8 mm coverslips precoated with 1% Matrigel. RPMI media was replaced every

other day. Cells were patched from days 5 to 15 after thaw. Each cell in this study was

acquired from a different coverslip that had been independently cultured for >4 days

in a 24-well plate. Each cell was treated with drugs independently of all other cells in

the study to produce independent experimental samples. Cells from five different vials

were used in this study (see further details in the Data analysis and statistics section

below).

Perforated patch-clamp experiments were conducted using borosilicate glass

pipettes pulled to a resistance of 2-4 MΩusing a flaming/brown micropipette puller

(Model P-1000; Sutter Instrument, Novato, CA). The pipettes were filled with intra-

cellular solution containing 10 mM NaCl, 130 mM KCl, 1 mM MgCl2, 10 mM CaCl2,

5.5 mM dextrose, 10 mM HEPES. Amphotericin B was used as the perforating agent.

Amphotericin B allows only monovalent ions to pass through the cell membrane, so a

high intrapipette calcium concentration was used as a quality control step. In the case

of an unintended rupture, calcium will diffuse into the cell and cause calcium-induced

toxicity, which will terminate the experiment. The pipette tip was first dipped into in-

tracellular solution with no amphotericin B for 2-5 s. Then, the pipette was backfilled

55



with intracellular solution containing 0.44 mM amphotericin B. The coverslips contain-

ing iPSC-CMs were placed in the bath and constantly perfused with an extracellular

solution at 35-37°C contain