Experiment Simulation Simulated Micro-Electrode Array ... · important stage of the drug discovery...
1
Simulated Micro-Electrode Array Recordings from Stem Cell-Derived Cardiomyocytes Louise A. Bowler 1 , Kate Harris 2 , David J. Gavaghan 1 and Gary R. Mirams 1 1 Computational Biology Group, Department of Computer Science, University of Oxford 2 Safety Assessment, GlaxoSmithKline, Ware, UK Introduction Testing a novel pharmaceutical compound for adverse cardiac side-effects is a vitally important stage of the drug discovery process. Human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) are a promising cell type for cardiac safety testing, but differences between these cells and those in the adult heart mean that results must be interpreted with care. The micro-electrode array (MEA) is a particularly interesting safety test because it demonstrates how tissue, rather than a single cell, responds to the presence of a compound. Can models of the micro-electrode array help interpret the experimental results? Perform MEA experiment with new compound Ion channel screening data (IC 50 values) Simulate MEA with proposed effects Do experimental and simulated biomarkers agree? Have reasonable understanding of compound’s effects on tissue Compound must affect other ion channels Preliminary safety assessment: terminate or proceed with further tests Yes No Simulation Design Simulating the Field Potential • The bidomain equations describe how the membrane and extracellular potentials vary over cardiac tissue. The equations were solved using Chaste, a C ++ library with a finite element solver for cardiac electrophysiology. • The ionic current was calculated using a cellular electrophysiology model: in this case, the atrial-like and ventricular-like hiPSC-CM models of Paci et al. (2013). • Biomarkers of interest such as the beat rate and field potential duration were derived from the simulated extracellular potential. Convergence Analysis • A square mesh (the same diameter of the MEA) was created with regularly-spaced nodes. • The space between the nodes and the time-steps used to solve the system were varied. • Space and time-steps were chosen so that reducing either of them further had no appreciable effect on the field potential or the values of the biomarkers. 0 2 4 6 8 10 -3 -2 -1 0 1 2 time / s Φ e / mV 500 250 125 62.5 31.25 15.625 8 8.5 9 9.5 10 -3 -2 -1 0 1 2 time / s Φ e / mV Above: Extracellular potential from simulations using different values of inter-node spacing (in μm). MEA Simulations • A 2D mesh was generated to model the monolayer of cells within the MEA, with maximum inter-node spacing of 150 μm. The time-step was 0.05 ms. • 20s of activity was simulated at each drug concentration. • The extracellular potential was recorded at the locations of the 60 electrodes. Far left: hiPSC-CM monolayer on an MEA. Left: Nodes in an MEA mesh, showing the atrial-like (red) and ventricular-like (blue) cellular electro- physiology models. In this example, the inter-node spacing is 250 μm. Ion Channel Block in the MEA The effects of ion channel blocking drugs on the simulated MEA were studied by altering the conductances of the relevant channels. The degree of block at each drug concentration (given in nM) was calculated from the IC 50 values (Mirams et al., 2011). Biomarkers from the simulated data were compared to experimental results (Harris et al., 2013). 0 0.5 1 1.5 2 -2 -1 0 1 time / s Φ e / mV RR interval FPD amplitude Left: Example of a typical simulated extracellular potential, labelled with the biomarkers used in this analysis. Corrected field potential duration (FPDc) was calculated from the FPD using Fridericia’s formula. Cisapride: hERG channel blocker • The change in FPDc is larger in the simulated results than in the experimental results. • Arrhythmias were reported in experiments at the highest cisapride concentration (1 μM). • Cells in other batches were reported to show a stronger response to similar hERG blockers than those shown here: low I Kr in this batch of cells could have affected the results. 0 0.2 0.4 0.6 0.8 -0.2 -0.1 0 0.1 0.2 time / s Φ e / mV 0 3 10 100 1000 3 10 100 1000 0 20 40 60 [Cisapride] / nM % change in FPDc Arrhythmias Simulation Experiment Nifedipine: Ca 2+ channel blocker • The change in FPDc agrees with the experimental results. • The form of the potential is distorted at the highest concentration of Nifedipine. 0 0.2 0.4 0.6 0.8 -0.2 -0.1 0 0.1 0.2 time / s Φ e / mV 0 30 100 300 1000 30 100 300 1000 -80 -60 -40 -20 0 [Nifedipine] / nM % change in FPDc Simulation Experiment Terfenadine: mixed channel blocker • The simulated results follow the trend of the experimental ones, with an enhanced effect. • As Terfenadine mainly blocks hERG at lower concentrations, it is possible that differences in I Kr in control conditions are responsible for some of the differences. • The reduction in FPDc at high concentrations is captured well by the simulated system. 0 0.2 0.4 0.6 0.8 -0.2 -0.1 0 0.1 0.2 time / s Φ e / mV 0 3 10 100 1000 3 10 100 1000 -20 -10 0 10 20 30 [Terfenadine] / nM % change in FPDc Simulation Experiment Simulations of the MEA can reproduce the observed trends in corrected field potential duration, but do not yet consistently provide quantitative predictions of drug effects. Future Work • Investigate alternative ways of representing regions of atrial-like and ventricular-like phenotype. This would include different sizes of atrial-like or ventricular-like regions, and the use of a mixed atrial-ventricular model for some or all of the simulated tissue. • Repeat the simulations using IC 50 values from alternative sources, such as high- throughput screening data. • Identify aspects of the electrophysiology models which could be improved, and consider whether the models should be re-parametrised for a given batch of cells. Acknowledgements References Contact Doctoral Training Centre Life Sciences Interface L S I Harris, K., M. Aylott, Y. Cui, J. B. Louttit, N. C. McMahon, and A. Sridhar (2013). Toxicol. Sci. 134(2): 412–426. Mirams, G. R., Y. Cui, A. Sher, M. Fink, J. Cooper, B. M. Heath, N. C. McMahon, D. J. Gavaghan, and D. Noble (2011). Cardiovasc. Res. 91(1): 53–61. Paci, M., J. Hyttinen, K. Aalto-Set¨ al¨ a, and S. Severi (2013). Ann. Biomed. Eng. 41(11): 2334–2348. If you have any questions about this poster, please ask me or get in touch via email: [email protected]
Transcript of Experiment Simulation Simulated Micro-Electrode Array ... · important stage of the drug discovery...
Simulated Micro-Electrode Array Recordingsfrom Stem Cell-Derived Cardiomyocytes
Louise A. Bowler1, Kate Harris2, David J. Gavaghan1 and Gary R. Mirams1
1Computational Biology Group, Department of Computer Science, University of Oxford2Safety Assessment, GlaxoSmithKline, Ware, UK
Simulated Micro-Electrode Array Recordingsfrom Stem Cell-Derived Cardiomyocytes
Louise A. Bowler1, Kate Harris2, David J. Gavaghan1 and Gary R. Mirams1
1Computational Biology Group, Department of Computer Science, University of Oxford2Safety Assessment, GlaxoSmithKline, Ware, UK
Introduction
Testing a novel pharmaceutical compound for adverse cardiac side-effects is a vitallyimportant stage of the drug discovery process.
Human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) are a promisingcell type for cardiac safety testing, but differences between these cells and those in the adultheart mean that results must be interpreted with care.
The micro-electrode array (MEA) is a particularly interesting safety test because itdemonstrates how tissue, rather than a single cell, responds to the presence of a compound.
Can models of the micro-electrode array help interpret the experimental results?
Perform MEAexperiment withnew compound
Ion channelscreening data(IC50 values)
Simulate MEA withproposed effects
Do experimentaland simulated
biomarkers agree?
Have reasonableunderstandingof compound’seffects on tissue
Compound mustaffect otherion channels
Preliminarysafety assessment:
terminate orproceed withfurther tests
Yes
No
Simulation Design
Simulating the Field Potential
• The bidomain equations describe how the membrane and extracellular potentials varyover cardiac tissue. The equations were solved using Chaste, a C++ library with a finiteelement solver for cardiac electrophysiology.
• The ionic current was calculated using a cellular electrophysiology model: in this case,the atrial-like and ventricular-like hiPSC-CM models of Paci et al. (2013).
• Biomarkers of interest such as the beat rate and field potential duration were derivedfrom the simulated extracellular potential.
Convergence Analysis
• A square mesh (the same diameter of the MEA) was created with regularly-spaced nodes.
• The space between the nodes and the time-steps used to solve the system were varied.
• Space and time-steps were chosen so that reducing either of them further had noappreciable effect on the field potential or the values of the biomarkers.
0 2 4 6 8 10−3
−2
−1
0
1
2
time / s
Φe /
mV
500 250 125 62.5 31.2515.625
8 8.5 9 9.5 10−3
−2
−1
0
1
2
time / s
Φe /
mV
500 250 125 62.5 31.2515.625
Above: Extracellular potential from simulations using different values of inter-node spacing (in µm).
MEA Simulations
• A 2D mesh was generated to model the monolayer of cells within the MEA, withmaximum inter-node spacing of 150µm. The time-step was 0.05 ms.
• 20 s of activity was simulated at each drug concentration.
• The extracellular potential was recorded at the locations of the 60 electrodes.
Far left: hiPSC-CM monolayer onan MEA.
Left: Nodes in an MEA mesh,showing the atrial-like (red) andventricular-like (blue) cellular electro-physiology models. In this example,the inter-node spacing is 250µm.
Ion Channel Block in the MEA
The effects of ion channel blocking drugs on the simulated MEA were studied by alteringthe conductances of the relevant channels. The degree of block at each drug concentration(given in nM) was calculated from the IC50 values (Mirams et al., 2011). Biomarkers fromthe simulated data were compared to experimental results (Harris et al., 2013).
0 0.5 1 1.5 2
−2
−1
0
1
time / s
Φe /
mV
RR interval
FPD
amplitudeLeft: Example of a typical simulatedextracellular potential, labelled with thebiomarkers used in this analysis. Correctedfield potential duration (FPDc) was calculatedfrom the FPD using Fridericia’s formula.
Cisapride: hERG channel blocker
• The change in FPDc is larger in the simulated results than in the experimental results.
• Arrhythmias were reported in experiments at the highest cisapride concentration (1 µM).
• Cells in other batches were reported to show a stronger response to similar hERG blockersthan those shown here: low IKr in this batch of cells could have affected the results.
0 0.2 0.4 0.6 0.8−0.2
−0.1
0
0.1
0.2
time / s
Φe /
mV
0 3 10 1001000
3 10 100 1000
0
20
40
60
[Cisapride] / nM
% c
hang
e in
FP
Dc
Arr
hyt
hm
ias
SimulationExperiment
Nifedipine: Ca2+ channel blocker
• The change in FPDc agrees with the experimental results.
• The form of the potential is distorted at the highest concentration of Nifedipine.
0 0.2 0.4 0.6 0.8−0.2
−0.1
0
0.1
0.2
time / s
Φe /
mV
0 30 100 3001000
30 100 300 1000−80
−60
−40
−20
0
[Nifedipine] / nM%
cha
nge
in F
PD
c
SimulationExperiment
Terfenadine: mixed channel blocker
• The simulated results follow the trend of the experimental ones, with an enhanced effect.
• As Terfenadine mainly blocks hERG at lower concentrations, it is possible that differencesin IKr in control conditions are responsible for some of the differences.
• The reduction in FPDc at high concentrations is captured well by the simulated system.
0 0.2 0.4 0.6 0.8−0.2
−0.1
0
0.1
0.2
time / s
Φe /
mV
0 3 10 1001000
3 10 100 1000−20
−10
0
10
20
30
[Terfenadine] / nM
% c
hang
e in
FP
Dc
SimulationExperiment
Simulations of the MEA can reproduce the observed trends in corrected field potentialduration, but do not yet consistently provide quantitative predictions of drug effects.
Future Work
• Investigate alternative ways of representing regions of atrial-like and ventricular-likephenotype. This would include different sizes of atrial-like or ventricular-like regions,and the use of a mixed atrial-ventricular model for some or all of the simulated tissue.
• Repeat the simulations using IC50 values from alternative sources, such as high-throughput screening data.
• Identify aspects of the electrophysiology models which could be improved, and considerwhether the models should be re-parametrised for a given batch of cells.
Acknowledgements References Contact
Doctoral Training CentreLife Sciences Interface L S
I
CASE STUDY
OverviewThe University of Oxford, renowned for its academic excellence, wanted to better support researchers from all faculties. With its existing cluster struggling to cope with growing demand for high performance computing, the university worked with Lenovo and OCF to deploy a Lenovo NeXtScale System M5 cluster with over 5,000 Intel processor cores. With the Lenovo solution, the university can now offer researchers from a wide range of disciplines a truly state-of-the-art research tool, helping to maintain its sparkling reputation.
The University of Oxford is a world-leading collegiate research university. Located in
Oxford, England, the university offers undergraduate and postgraduate degrees to
over 22,000 students, and consistently ranks at the top of international higher
education league tables.
Dr. Andrew Richards, Head of Advanced Research Computing at the University of
Oxford, says: “The university is one of the highest-ranking academic institutions in
the world, renowned for the quality of its facilities, research and teaching. In recent
years, demand for high performance computing (HPC) resources has been growing
steadily. To adequately support a wide range of research projects, we knew we
needed to boost the capacity and processing power of our existing clusters.”
High demand for high performanceThe Advanced Research Computing facility at the University of Oxford is a central
resource available to all university-affiliated researchers in need of high performance
computing (HPC). Traditionally the preserve of physical, medical and biological
sciences research, HPC systems are now frequently used for computational analysis
and data modeling across a wide range of academic disciplines, including humanities
and social sciences.
“The Lenovo supercomputer is helping the University of Oxford maintain its excellent reputation for research.”—Dr. Andrew Richards, Head of Advanced Research Computing, University of Oxford
University of Oxford supports world-class academic research with Lenovo supercomputer
EDUCATION
Harris, K., M. Aylott, Y. Cui, J. B. Louttit, N. C. McMahon, and A. Sridhar (2013). Toxicol. Sci.134(2): 412–426.
Mirams, G. R., Y. Cui, A. Sher, M. Fink, J. Cooper, B. M. Heath, N. C. McMahon, D. J. Gavaghan,and D. Noble (2011). Cardiovasc. Res. 91(1): 53–61.
Paci, M., J. Hyttinen, K. Aalto-Setala, and S. Severi (2013). Ann. Biomed. Eng. 41(11): 2334–2348.
If you have any questions aboutthis poster, please ask me or getin touch via email:
![Human-based approaches to pharmacology and cardiology: an ... · induced pluripotent stem cells (hiPSC-CMs) [Daniels, Severi, Kopljar, Harmer] – Inconsistent immaturity; – Variability](https://static.fdocuments.in/doc/165x107/5fb2c5171f4b03320f31801d/human-based-approaches-to-pharmacology-and-cardiology-an-induced-pluripotent.jpg)
