Consistent Analysis of Action Potential Parameters for Various Types of Cardiomyocytes

Introduction
Analyzing cardiac action potential (AP) properties has been a key method in arrhythmia and drug development research for decades. Recently, human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) have become an emerging tool in cardiovascular research. The spontaneous electrical activity of hiPSC-CMs allows simultaneous investigation of parameters from both individual APs and the collective recording of multiple APs such as beating frequency or arrhythmogenicity in vitro. However, due to the diverse morphologies of hiPSC-CM APs, manual extraction of standard AP parameters is time-consuming and depends on the operator’s subjective judgment which hampers reproducible, quantitative comparisons between various types of CMs. Here, we present a consistent, fast, reproducible and operator-independent algorithm to automatically extract quantitative AP parameters from recordings of various CM types.

Methods
Due to the variability and noise in hiPSC-CM AP recordings, classical mathematical tools or thresholding cannot be directly applied. Instead, APs are identified as sequences characterized by a potential rise exceeding 20 mV within a time frame of 25 ms. Markers for the start of an AP, maximum upstroke velocity, peak potential and maximum diastolic potential (MDP) are derived automatically (for further details, see https://gitlab.kit.edu/kit/ibt-public/analysis-of-ap-parameters). The analyzed AP parameters are illustrated in Figure 1, complemented by the frequency and action potential duration (APD) corrected by Bazett’s formula for varying beating frequencies.

Results
The algorithm was tested on 501 AP recordings from hiPSC-CMs, 873 AP recordings from stimulated isolated native CMs, 133 AP recordings from HL-1 cells and 100 AP recordings from computational electrophysiological simulations. In the absence of a ground truth, the algorithm was validated through comprehensive manual expert inspection. The markers placed by the algorithm were considered correct in over 99% of cases, computed in 0.4 s for a 60 s AP recording containing 20 APs sampled at 10 kHz.
The tested AP recordings exhibited substantial variability, with maximum upstroke velocities ranging from 2 to 120 mV/ms, MDPs from -82 to -40 mV, peak potentials from 20 to 50 mV, APD₉₀ from 50 to 700 ms and beating frequencies from 0.5 to 4 Hz.
Exemplarily, we analyzed AP parameters in 12 isolated native control CMs and 16 atrial-like control hiPSC-CMs. The spontaneously beating hiPSC-CMs showed a more depolarized MDP (-49.17 ± 6.93 mV vs. -68.86 ± 2.27 mV, all values are mean ± SD) and thus a lower maximum upstroke velocity (9.78 ± 11.78 mV/ms vs. 56.73 ± 12.15 mV/ms) and AP amplitude (75.79 ± 13.22 mV vs. 110.60 ± 12.23 mV) compared to isolated native CMs. Conversely, the APD₉₀ and area of APD₉₀ were greater in hiPSC-CMs compared to isolated native CMs (276.77 ± 57.08 ms vs. 89.76 ± 31.14 ms and 7.50 ± 2.72 mV*s vs. 2.08 ± 0.83 mV*s).

Conclusion
This algorithm enables high-throughput analyses of diverse AP recordings, facilitating quantitative comparisons of electrophysiological properties in various types of CMs.