Development, calibration, and validation of a novel human ventricular myocyte model in health, disease, and drug block

Figure 1.. Model structure. (A) A schematic of the novel human ventricular myocyte model for electrophysiology and calcium handling. Orange indicates components, substituted, or added, compared to the original ORd model. ‘SS’ indicates junctional subspace compartment, where calcium influx via L-type calcium current occurs and where calcium is released from the sarcoplasmic reticulum. ‘JSR’ and ‘NSR’ are junctional and network sarcoplasmic reticulum compartments, respectively. ‘Main cytosolic pool’ is the remaining intracellular space. Transmembrane currents are indicated with an ‘I’ in their name, with fluxes indicated as ‘J’. Components with a green underscore are modulated by CaMKII signalling. (B) The structure of the Lu-Vandenberg (Lu et al., 2001) Markov model used for the rapidly activating delayed rectifier repolarisation current (IKr). The transition rates are given in Appendix 1-15.3.5.

Figure 2.. Action potential, calcium transient, and ICaL in ToR-ORd. Action potential (A) and calcium transient (B) at 1 Hz obtained with the ToR-ORd model following calibration, compared to those obtained with the ORd model and experimental data from O'Hara et al. (2011) and Coppini et al. (2013), respectively. The purple and green zones in (B) stand for mean ± standard deviation. The duration of calcium transient at 90% recovery was extracted from figures in Coppini et al. (2013), adding the time to peak and time from peak to 90% recovery. (C) Activation curves used in the ToR-ORd and ORd models (blue and red lines, respectively). The points correspond to the IV relationship measured in Magyar et al. (2000) normalised by the Nernstian driving force (d.f.) assuming reversal potential of +60 mV (blue points) and the GHK driving force (red points). (D, E) I-V relationship and steady-state inactivation as measured in ToR-ORd (red line) versus ORd model (blue line) versus experimental data from Magyar et al. (2000) (black points with line). ICaL,tot is the sum of currents corresponding to all ions (Ca2+, Na+, and K+) passing through the L-type calcium current channels. (F) L-type calcium…

Figure 3.. Sodium blockade in ToR-ORd and ORd. Simulated effect of sodium current block on the action potential (A, B) and calcium transient (C,D) using the ToR-ORd (left) and ORd (right) models. Control simulations are shown as blue traces, whereas results for 50% block of INa and INaL are shown as red traces. Panels (E,F) show values of changes in calcium transient amplitude with respect to control (‘rel. CaTA’) for varying degrees of INa/INaL block for ToR-ORd and ORd respectively (one is no change, values > 1 correspond to an increase the calcium transient amplitude).

Figure 4.. EADs and alternans. (A) Experimentally observed EADs produced in nondiseased human myocytes exposed to 0.1 µM dofetilide (∼85% IKr block) paced at 0.25 Hz (Guo et al., 2011) and the corresponding simulation using the ORd model (O'Hara et al., 2011), figure reproduced as allowed by the CC-BY license). See Appendix 1-15.1.1 for ionic concentrations used. (B) An EAD produced by the ToR-ORd model at the same conditions. (C,E) APD and calcium transient amplitude versus pacing base cycle length (bcl); the bifurcation indicates alternans. (D,F) Example of APD and calcium transient alternans at the bcl of 260 ms.

Figure 5.. Drug block and APD. All four panels contain mean and standard deviation of experimental data (black) as measured in O'Hara et al. (2011) for three basic cycle lengths (bcl), predictions of the Dutta optimised dynamic-IKr ORd model (blue, Dutta et al., 2017a), and the predictions of the ToR-ORd model (red). (A) 1 µM E-4031 (70% IKr block). (B) 1 µM HMR-1556 (90% IKs block). (C) 10 µM mexiletine (54% INaL, 9% IKr, 20% ICaL block). (D) 1 µM nisoldipine (90% ICaL block). Drug concentrations and their effects on channel blocks are based on Dutta et al. (2017b).

Figure 6.. APD accommodation and S1-S2 protocol. (A) APD accommodation measured experimentally (Franz et al., 1988) and in simulation of the ORd model (reproduced as allowed by the CC-BY licence from O'Hara et al., 2011). (B) APD accommodation in ToR-ORd. (C) S1-S2 restitution curve (S1 = 1000 ms) in ToR-ORd fibre, ORd fibre (including INa modification to facilitate propagation Passini et al., 2016) and human tissue samples data (O'Hara et al., 2011).

Figure 7.. Populations of models and drug safety prediction. (A) Percentile-based summary of AP and calcium transient traces for the two populations of human ToR-ORd models (left side) and distribution of ionic current conductances among models in the population (right side), in the ranges [50-150]% (top row) and [0–200]% (bottom row) of the baseline values. (B) A comparison of 0–200% populations based on ToR-ORd (left) and ORd (right, based on Passini et al., 2017) in response to high-dose mexiletine (100-fold effective therapeutic dose). Traces classified as EADs are plotted in red (manifesting only in the ORd population). (C) TdP score obtained for simulations of the 62 reference compounds, based on the occurrence of drug-induced repolarisation abnormalities at all tested concentrations in the [0–200]% population of ToR-ORd models. The colours associated with drugs signify their established torsadogenicity as specified in Appendix 1-15.1.4. The logarithmic scale was considered to maximise the visual separation between safe and risky drugs. The classification based on the TdP score is summarised as a confusion matrix on the right, and also compared with the corresponding results obtained…

Figure 8.. Simulation of hypertrophic cardiomyopathy (HCM) and hyperkalemia. (A) The effect of hyperkalemia on AP morphology; measured in the centre of a simulated fibre. (B) APD90 and effective refractory period (ERP) at varying extracellular potassium concentration. For extracellular potassium higher than 9 mM, full AP did not develop, but low-amplitude activation propagated through the fibre (Appendix 1-14). Membrane potential (C) and calcium transient (D) at 1 Hz pacing compared between a single healthy and HCM cell. (E) 50% IKr block induces EADs in HCM cell, but not in a healthy one.

Figure 9.. Simulated and clinical 12-lead electrocardiogram. (A) 12-lead ECGs at 1 Hz: simulation using the ToR-ORd model in an MRI-based human torso-ventricular model (top) and a healthy patient ECG record (bottom, https://physionet.org, PTB database, subject 122; Bousseljot et al., 1995; Goldberger et al., 2000). (B) Electrode positions on the simulated torso. (C) Activation time map. (D) APD map.

Appendix 1—figure 1.. Diagram of how activation curves are extracted from the I-V relationship. The activation curves (top right) are obtained by dividing values in the I-V curve (top left) at each pulse potential by the driving force considering either V-ECa-based driving force (bottom right) or the Goldman-Hodgkin-Katz (GHK) flux equation (bottom left).

Appendix 1—figure 2.. Effect of the proposed changes in the activation curve and activation coefficient on the I-V relationship of ICaL in ORd versus ToR-ORd.

Appendix 1—figure 3.. Goldman-Hodgkin-Katz equation with colour coding of components.