Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Cardiac strength-interval curves calculated using a bidomain tissue with a parsimonious ionic current

Cardiac strength-interval curves calculated using a bidomain tissue with a parsimonious ionic current

  • Suran K. Galappaththige, 
  • Richard A. Gray, 
  • Bradley J. Roth
PLOS
x

Abstract

The strength-interval curve plays a major role in understanding how cardiac tissue responds to an electrical stimulus. This complex behavior has been studied previously using the bidomain formulation incorporating the Beeler-Reuter and Luo-Rudy dynamic ionic current models. The complexity of these models renders the interpretation and extrapolation of simulation results problematic. Here we utilize a recently developed parsimonious ionic current model with only two currents—a sodium current that activates rapidly upon depolarization INa and a time-independent inwardly rectifying repolarization current IK—which reproduces many experimentally measured action potential waveforms. Bidomain tissue simulations with this ionic current model reproduce the distinctive dip in the anodal (but not cathodal) strength-interval curve. Studying model variants elucidates the necessary and sufficient physiological conditions to predict the polarity dependent dip: a voltage and time dependent INa, a nonlinear rectifying repolarization current, and bidomain tissue with unequal anisotropy ratios.

Introduction

An in-depth understanding of cardiac excitability is critical to the development of pacemakers and defibrillators. Modeling the dynamics of wave excitation in cardiac tissue is a way to predict the complex behavior of the tissue in response to an applied stimulus. Many models have been used to predict this behavior; typically these models have numerous parameters and ion currents, and each model has its own strengths and weaknesses [1]. The complexity of these models make it difficult to identify the relative role of parameters and variables on cell behavior such as the action potential duration (APD) and the excitation threshold. Other “phenomenological” models are simpler with fewer parameters, but are less realistic, and do not allow a link to detailed physiology such as voltage clamp experiments.

The strength-interval (SI) curve is a useful way to characterize the behavior of cardiac tissue in response to a stimulus [2]. Two stimuli are applied separated by an interval of time. The first (S1) triggers an action potential, and the second (S2) probes the tissue excitability as the tissue recovers from S1 refractoriness. Previous experiments [3], as well as simulations using the bidomain model [4] with the Beeler-Reuter model [5], have shown the mechanisms of excitation underlying cathodal and anodal SI curves are cathode make, cathode break, anode make, and anode break. “Cathode” and “anode” refer to the polarity of the stimulus electrode, and “make” and “break” to the mechanism of tissue excitation; make is when excitation begins directly from depolarized tissue following the onset of the stimulus pulse, and break is when it begins (usually indirectly through hyperpolarized tissue) following the termination of the pulse.

Similar modeling studies [6], including one using the Luo-Rudy dynamic model [7, 8], analyzed the “dip” in the anodal SI curve. Typically, the SI curve has negative slope, indicating that the S2 threshold is lower at longer intervals when the tissue has recovered from S1 refractoriness. For anodal stimulation, however, part of the SI curve has a positive slope, indicating that S2 excitation is easier when the tissue is less recovered. This counterintuitive behavior is prominent in the break section of the anodal SI curve, and is suppressed or absent from the cathodal SI curve. One proposed mechanism responsible for the dip is that following the S2 stimulus, the tissue surrounding the anode is hyperpolarized and excitable. The limiting factor determining the threshold is how well this excitable tissue interacts with nearby depolarization that can be produced by either the S2 stimulus or the previous S1 action potential. As the interval is shortened, more S1 depolarization is present so less S2 depolarization is required, lowering the threshold S2 stimulus strength. Thus, the dip arises from the electrotonic interaction of excitable tissue with surrounding depolarization [2,6,7]. Other studies have claimed that the mechanism of the dip is caused by calcium dynamics [9] or is an intrinsic property of the membrane current [10,11].

A recently developed parsimonious ionic model [12, 13] reproduces many experimentally measured characteristics of ventricular action potentials using only three variables, two ion currents, and eleven parameters. The advantage of a parsimonious model is that it highlights essential features of the membrane kinetics while not getting bogged down in nonessential details. The purpose of this manuscript is to combine the parsimonious ionic model with the bidomain formulation to study the events responsible for the shape of the strength interval curve. Specifically, we explore which features of the parsimonious model are essential for predicting the anomalous dip in the anodal SI curve. Our hypothesis is that a parsimonious model coupled with the bidomain formulation is sufficient to predict the shapes of the anodal and cathodal SI curves, and that the nonlinearity of the repolarization current is necessary for the dip in the anodal SI curve.

Results

Space-Clamped model

Anodal stimulation (hyperpolarization) did not cause excitation in the space-clamped parsimonious model. Space-clamped cathodal stimulation (depolarization) resulted in supernormal excitation immediately following a previous S1 action potential, which may arise because the transmembrane potential has not yet returned to rest and therefore less depolarization is required to reach threshold, even though INa is not fully inactivated [14, 15].

Parsimonious ionic current model with bidomain tissue model

The strength interval curves for both cathodal and anodal stimuli applied through a unipolar electrode for a 2 ms pulse are shown in Fig 1, using the standard model parameters [12, 13]. The S1-S2 interval is defined as the time between the start of the S1 stimulus pulse (at time 0) and the start of the S2 pulse. These tissue simulations using the bidomain model with unequal anisotropy ratios predict a complicated transmembrane potential distribution with adjacent regions of depolarization and hyperpolarization [3,16] for both anodal and cathodal stimulation.

thumbnail
Fig 1. The cathodal and anodal strength-interval curves for the parsimonious ionic model.

Cathodal (blue) and anodal (red), with kr = 21.28 mV, gK = 0.3 mS cm-2, and EK = -83 mV. The simulations included as movies in the supplementary information are indicated by black open ovals.

http://dx.doi.org/10.1371/journal.pone.0171144.g001

Cathodal SI curve.

The SI curve for cathodal stimulation (Fig 1, blue curve) shows a decrease in threshold S2 stimulus strength as the tissue is recovering from the S1 action potential. At long intervals (rheobase), the cathodal threshold is 30 mA/cm. At an interval of 155 ms the tissue is nearly recovered from S1 refractoriness and cathode make excitation originates from the depolarization surrounding the cathode. Details of this behavior are shown in Fig 2 and S1 Movie (interval = 155 ms, S2 stimulus = 33 mA/cm; see supplementary information). At shorter intervals cathode break excitation occurs. The tissue is still refractory at 143 ms when the S2 pulse begins. During and immediately after the S2 stimulus the depolarization is extensive and hyperpolarization along the fiber direction is evident (the “virtual anode”). Break excitation occurs following the end of the S2 pulse, when the depolarization surrounding the cathode diffuses into the hyperpolarized virtual anode, exciting a wave front that initially propagates through the hyperpolarized and therefore excitable tissue along the fiber axis (Fig 3 and S2 Movie, 143 ms, 435 mA/cm). At less than 140 ms, the tissue is unexcitable for any S2 strength (absolute refractory).

thumbnail
Fig 2. Cathode make excitation.

The transmembrane potential is shown as a function of x (horizontal, parallel to the fibers) and y (vertical, perpendicular to the fibers). The black rectangle is the electrode and the number in each frame is the time in ms. S1-S2 interval = 155 ms, S2 stimulus = 33 mA/cm. These frames are taken from the S1 Movie (see supplementary files). The color scale for this and other plots of transmembrane potential is shown in the movie; purple is rest, yellow is depolarized, and blue is hyperpolarized.

http://dx.doi.org/10.1371/journal.pone.0171144.g002

thumbnail
Fig 3. Cathode break excitation.

The transmembrane potential is shown as a function of x (horizontal, parallel to the fibers) and y (vertical, perpendicular to the fibers). The black rectangle is the electrode and the number in each frame is the time in ms. S1-S2 interval = 143 ms, S2 stimulus = 435 mA/cm. These frames are taken from the S2 Movie (see supplementary files).

http://dx.doi.org/10.1371/journal.pone.0171144.g003

Anodal SI curve.

The anodal SI curve (Fig 1, red curve) has a prominent dip (a local minimum created by an abrupt drop at 127 ms followed by a region of positive slope), as seen in other models [2,6,7]. The threshold value of the anodal stimulus at the minimum of the dip is 587 mA/cm. For S1-S2 intervals below 127 ms the tissue is unexcitable. The maximum threshold value is 1017 mA/cm recorded at 147 ms. The range from 155 to 160 ms contains a supernormal period similar to that in the space-clamped simulations, during which the threshold dips to at most 3% below its asymptotic long-interval (rheobase) value of 598 mA/cm.

Anode make excitation occurs at long intervals. The hyperpolarization surrounding the anode plays little role in excitation, and the wave front originates from the depolarized region (the “virtual cathode”) along the fiber direction (Fig 4 and S3 Movie, 158 ms, 640 mA/cm). The supernormal period (155–160 ms) was restricted to the make section of the SI curve. Anode break excitation occurs at shorter intervals. The depolarization along the fiber axis diffuses into the hyperpolarized region surrounding the anode, exciting a wave front that propagates initially perpendicular to the fiber direction (Fig 5 and S4 Movie, 130 ms, 710 mA/cm). At intervals shorter than 127 ms, anode break initiates an action potential but it can only propagate locally near the electrode because the surrounding tissue is too refractory and therefore is incapable of generating a wave that propagates through the heart (Fig 6 and S5 Movie, 125 ms, 600 mA/cm).

thumbnail
Fig 4. Anode make excitation.

The transmembrane potential is shown as a function of x (horizontal, parallel to the fibers) and y (vertical, perpendicular to the fibers). The black rectangle is the electrode and the number in each frame is the time in ms. S1-S2 interval = 158 ms, S2 stimulus = 640 mA/cm. These frames are taken from the S3 Movie (see supplementary files).

http://dx.doi.org/10.1371/journal.pone.0171144.g004

thumbnail
Fig 5. Anode break excitation.

The transmembrane potential is shown as a function of x (horizontal, parallel to the fibers) and y (vertical, perpendicular to the fibers). The black rectangle is the electrode and the number in each frame is the time in ms. S1-S2 interval = 130 ms, S2 stimulus = 710 mA/cm. These frames are taken from the S4 Movie (see supplementary files).

http://dx.doi.org/10.1371/journal.pone.0171144.g005

thumbnail
Fig 6. Failed anode break excitation.

The transmembrane potential is shown as a function of x (horizontal, parallel to the fibers) and y (vertical, perpendicular to the fibers). The black rectangle is the electrode and the number in each frame is the time in ms. S1-S2 interval = 125 ms, S2 stimulus = 600 mA/cm. These frames are taken from S5 Movie (see supplementary files).

http://dx.doi.org/10.1371/journal.pone.0171144.g006

Equal anisotropy ratios

The SI curves were recalculated using equal anisotropy ratios, so no virtual anode was produced during cathodal stimulation and no virtual cathode occurred during anodal stimulation. Anodal stimulation did not excite the tissue. Cathodal stimulation was always by the make mechanism and not the break mechanism.

Effect of IK current on strength-interval curves

The repolarization current IK contains three parameters (see Methods section), each of which modifies the IK current-voltage curve in a unique way. The conductance gK scales the magnitude of the IK current, EK is the reversal potential, and kr controls the nonlinearity of the current-voltage curve.

The variation of IK with transmembrane potential for different kr values is shown in Fig 7a. The values for kr are chosen to show the effect of the IK current when it is linear (kr = ∞) and nonlinear (kr = 21.28 and 40 mV). The value kr = 21.28 mV corresponds to that used in Gray’s original model [12, 13]. When gK and EK are fixed (0.3 mS cm-2 and -83 mV), the current-voltage curves have the same intercept and slope for IK = 0.

thumbnail
Fig 7. Repolarization current and transmembrane potential for different kr values.

(a) IK vs Vm and (b) Vm vs time at x = 0.35 cm and y = 0.1 cm, for different kr values. The parameters producing the red, purple, and black curves were chosen so the action potential duration is approximately the same.

http://dx.doi.org/10.1371/journal.pone.0171144.g007

Cathodal SI curves.

The calculated SI curves are shown in Fig 8. All cathodal SI curves show a decrease in the S2 threshold strength as the S1-S2 interval increases [2]. The cathodal rheobase varies little with kr; for kr = 21.28 mV the S2 threshold is 33 mA/cm, for kr = 40 mV S2 is 30 mA/cm, and for kr = ∞ S2 is 31 mA/cm. The cathodal SI curves show the transition from break (solid symbols) to make (open symbols); for kr = ∞ the transition occurs at 15 ms, for kr = 40 mV it occurs at 37 ms, and for kr = 21.28 mV it occurs at 155 ms. The shift of the SI curve to shorter intervals as kr increases reflects the time taken by the tissue to recover from refractoriness, where a change in kr changes the duration of the propagating action potential (Fig 7b). For the cathodal SI curve with kr = 21.28 mV the tissue is unexcitable below 140 ms, for kr = 40 mV below 27 ms, and for kr = ∞ below 11 ms.

thumbnail
Fig 8. Comparison of SI curves for different kr values.

kr = 21.28 mV red, kr = 40 mV blue, and kr = ∞ green. CM = cathode make (open circles), CB = cathode break (filled circles), AM = anode make (open squares), and AB = anode break (filled squares).

http://dx.doi.org/10.1371/journal.pone.0171144.g008

Anodal SI curves.

The IK current has more of an effect on the shape of the anodal SI curve than on the cathodal curve (Fig 8). For a linear IK current (kr = ∞) there is no anode break excitation and no dip in the anodal SI curve. The threshold strength for S2 remains almost constant with an increase in S1-S2 interval and the rheobase strength is highest of all kr values examined.

The APD and the amplitude of the action potential depend upon kr. With a nonlinear IK current the anodal SI curve exhibits a transition from break to make excitation; for kr = 21.28 mV the transition occurs at 155 ms, and for kr = 40 mV at 38 ms. The longer APD (kr = 21.28 mV) results in a SI curve with a smaller threshold dip than the short APD (kr = 40 mV). When kr = 40 mV, the anodal SI curve has a dip at 19 ms with a stimulus strength of 981 mA/cm. The maximum threshold current is 1565 mA/cm at 31 ms. For kr = 21.28 mV, anodal SI curve dip is at 127 ms with a strength of 587 mA/cm. Rheobase is highest for kr = ∞.

Effect of kr and gK parameters on strength-interval curves

One limitation of varying kr by itself is that the action potential duration changes dramatically. In order to look at changes in the nonlinearity of IK without confounding changes in APD, we varied both kr and gK simultaneously to keep the APD approximately constant. Fig 7b shows that kr = 40 mV and gK = 0.06 mS cm-2 produces a triangular shaped action potential, and that kr = ∞ mV and gK = 0.016 mS cm-2 produces a decaying exponential action potential, both with a duration similar to the standard action potential (kr = 21.28 mV and gK = 0.3 mS cm-2). Fig 7a shows the variation of the IK current with membrane potential. The SI curves of these three cases are shown in Fig 9. The anodal strength-interval curve for the linear case (kr = ∞) does not contain a section for anode break excitation, and does not contain a dip.

thumbnail
Fig 9. SI curves for different nonlinearities but similar action potential durations.

kr = 21.28 mV and gK = 0.3 mS cm-2 red, kr = 40 mV and gK = 0.06 mS cm-2 purple, and kr = ∞ and gK = 0.016 mS cm-2 black. The parameters were chosen so the action potential duration is approximately the same in each case. CM = cathode make (open circles), CB = cathode break (filled circles), AM = anode make (open squares), and AB = anode break (filled squares).

http://dx.doi.org/10.1371/journal.pone.0171144.g009

A nonlinear IK current is necessary for anode break excitation. To examine why, we performed simulations of anodal excitation for an interval of 130 ms, for kr = 21.28 mV and gK = 0.3 mS cm-2 (normal, nonlinear) and for kr = ∞ and gK = 0.016 mS cm-2 (linear). Fig 10 and S6 Movie show the normal case, and Fig 11 and S7 Movie show the linear case (both 130 ms, 750 mA/cm). Fig 12 shows the time course of the transmembrane potential at two locations, immediately adjacent to the anode (Fig 12a) and at the virtual cathode (Fig 12b). In the normal case, the hyperpolarization surrounding the anode is relatively weak and decays quickly after the stimulus pulse ends, creating an excitable path for the break wave front [5, 6, 17]. In the linear case, the hyperpolarization surrounding the anode is strong and decays slowly. The depolarization does not diffuse into hyperpolarized and excitable tissue. Rather, hyperpolarization diffuses into depolarized and refractory tissue, which does not result in excitation [18]. The nonlinearity of the repolarization current is critical to reducing the extent of the hyperpolarization.

thumbnail
Fig 10. Anode break excitation for the nonlinear case.

kr = 21.28 mV and gK = 0.3 mS cm-2 (normal, nonlinear). The transmembrane potential is shown as a function of x (horizontal, parallel to the fibers) and y (vertical, perpendicular to the fibers). The black rectangle is the electrode and the number in each frame is the time in ms. S1-S2 interval = 130 ms, S2 stimulus = 750 mA/cm. These frames are taken from S6 Movie (see supplementary files).

http://dx.doi.org/10.1371/journal.pone.0171144.g010

thumbnail
Fig 11. Failed anodal excitation for the linear case.

kr = ∞ and gK = 0.016 mS cm-2 (linear). The transmembrane potential is shown as a function of x (horizontal, parallel to the fibers) and y (vertical, perpendicular to the fibers). The black rectangle is the electrode and the number in each frame is the time in ms. S1-S2 interval = 130 ms, S2 stimulus = 750 mA/cm. These frames are taken from S7 Movie (see supplementary files).

http://dx.doi.org/10.1371/journal.pone.0171144.g011

thumbnail
Fig 12. The transmembrane potential as a function of time for nonlinear and linear cases.

The transmembrane potential at point A near the anode (x = 0.055 cm, y = 0.026 cm) and at point B in the virtual cathode (x = 0.255 cm, y = 0.026 cm), for a normal nonlinear (kr = 21.28 mV and gK = 0.3 mS cm-2) and a linear (kr = ∞ mV and gK = 0.016 mS cm-2) repolarization current. The S2 stimulus begins at 130 ms.

http://dx.doi.org/10.1371/journal.pone.0171144.g012

Effect of EK on strength-interval curve

At longer time intervals the primary mechanism of anodal excitation is make, whereas for shorter time intervals the primary mechanism is break [19]. These effects are evident in our calculations with the parsimonious ionic model (anode make in Fig 4 and anode break in Fig 5). When we increase the reversal potential (EK) to -69 mV, which represents an elevation in extracellular potassium ion concentration, at long time intervals the excitation mechanism is a combination of make and break (Fig 13 and S8 Movie, 200 ms, 500 mA/cm). Excitation with a weaker stimulus just above threshold (200 ms, 458 mA/cm) is purely break excitation, so we will categorize this more complex behavior as break. The change from make to break excitation at long intervals is consistent with the experimental work on elevated extracellular potassium ion concentration by Sidorov et al. [20], who showed that in diastole the anodal excitation changes from make to break. Numerical simulations carried out by Roth and Patel [21] using the Luo-Rudy dynamic model [8] suggest that potassium ion concentration influences anodal stimulation of cardiac tissue. Fig 14 shows the anodal and cathodal SI curves for EK = -83 and -69 mV. The action potential duration shortens with an increase in EK (Fig 15), causing the SI curves to shift to the left. Although the cathodal SI curve retains its shape the anodal SI curve for elevated potassium does not contain a dip. Therefore, elevated EK suppresses the dip in the anodal SI curve, consistent with previous studies [20, 21]. It also suppresses the supernormal period.

thumbnail
Fig 13. Anode make/break excitation for an elevated extracellular potassium ion concentration.

EK = -69 mV. The transmembrane potential is shown as a function of x (horizontal, parallel to the fibers) and y (vertical, perpendicular to the fibers). The black rectangle is the electrode and the number in each frame is the time in ms. S1-S2 interval = 200 ms, S2 stimulus = 458 mA/cm. These frames are taken from the S8 Movie (see supplementary files).

http://dx.doi.org/10.1371/journal.pone.0171144.g013

thumbnail
Fig 14. Comparison of SI curves for different EK values.

Normal (EK = -83 mV red) and elevated (EK = -69 mV blue) extracellular potassium concentration. CM = cathode make (open circles), CB = cathode break (filled circles), AM = anode make (open squares), and AB = anode break (filled squares).

http://dx.doi.org/10.1371/journal.pone.0171144.g014

thumbnail
Fig 15. Transmembrane potential for normal and elevated potassium.

Vm vs time at x = 0.35 cm and y = 0.1 cm for different EK values (-83 mV normal red, and -69 mV elevated potassium blue).

http://dx.doi.org/10.1371/journal.pone.0171144.g015

Discussion

Our simulations using the parsimonious ionic model for SI curves are consistent with results from previous studies [2,6,7,20,21]. The presence of the dip using a parsimonious model supports our hypothesis that the shape of the strength-interval curve is strongly influenced by the voltage dependent membrane recovery and bidomain properties of the tissue. By changing the parameter kr in the parsimonious ionic model we were able to obtain SI curves with shorter APD yet which have a transition of anode break excitation to anode make excitation (for kr = 21.28, 40 mV) as shown in movies in the supplementary files. The presence of the dip only at non-infinite values of kr suggests that a nonlinear rectifying repolarization current is necessary for anode break excitation and the dip to occur in the anodal SI curve. Moreover, the parsimonious ionic model predicts the effect of elevated extracellular potassium concentration. Roth and Patel [21] found that the disappearance of the dip was caused by post-repolarization refractoriness, so that when the tissue recovered from refractoriness there was no S1 depolarization to assist with anode break excitation.

Why did we choose to use a parsimonious model rather than a more detailed model for the ionic current? Gray and Pathmanathan [13] make a compelling case for the utility of minimal models that contain few parameters. Most more detailed ionic models are derived from voltage clamp data obtained under non-physiological conditions. Complex models suffer from over-parameterization, non-uniqueness, and have limited validity beyond the data used in their derivation. The parsimonious model is mathematically identifiable [12] and exhibits important emergent phenomena such as alternans and spiral wave breakup [13]. This model is attractive for us because we wish to know if the shape of the strength-interval curve can be predicted using even a simple model of cardiac electrical behavior. Our conclusion is that the parsimonious model can predict the shape of the SI curve, and that more complicated and exotic ionic currents are not necessary to explain this behavior.

The parsimonious ionic model has its limitations. APD depends upon the degree of linearity of IK. The ideal model is such that we should be able to vary the degree of linearity without a change in APD, so that we can isolate the effect of nonlinearity. There is a discrepancy between the quantitative value of the threshold currents for SI curves compared to previous studies. The currents are relatively large for the parsimonious ionic model. There can be numerous reasons for this discrepancy [22].

In conclusion, a simple ion current model that includes only an excitable sodium current and a nonlinear repolarization current is sufficient to predict the dip in the anodal SI curve. The parsimonious ionic model with the bidomain tissue model can predict shapes of the anodal and cathodal SI curves and break and make excitations for both anodal and cathodal stimuli. Nonlinearity of the repolarization current and electrotonic interaction of depolarization and hyperpolarization are both necessary for a dip in the anodal strength-interval curve. Other researchers have postulated that the dip may be caused by the “funny current” that activates upon hyperpolarization but that has a depolarized reversal potential so it supplies inward current near rest potential [10,11], or by calcium dynamics [9]. Although we cannot rule out a role for such currents with our analysis, this parsimonious model indicates that such factors are not necessary to reproduce the observed strength-interval curves.

Methods

The two dimensional bidomain model [23] is used to represent the cardiac tissue. The active membrane current is defined by the parsimonious ionic model [12, 13], which consists of two currents: a sodium current INa that activates rapidly upon depolarization, and a time-independent inward-rectifying repolarization current IK (1) (2) where (3) (4) with (5) (6) (7) (8)

The bidomain model is, (9) (10) where Vm and Ve are the transmembrane and extracellular potentials, Cm is the membrane capacitance per unit area, and β is the surface-to-volume ratio. Values for the intracellular and extracellular conductivities in the directions parallel and perpendicular to the fibers, gix, giy, gex, gey (Table 1), are chosen such that the tissue has unequal anisotropy ratios [16]. The stimulus current is applied at the center of the tissue using a rectangular unipolar electrode having a length (x-direction) and width (y-direction) of 0.1 cm. The stimulus is applied to a tissue of 1.0 cm by 1.0 cm by applying boundary conditions to the electrode tissue surface: we consider that the total current applied to the electrode is transferred to the extracellular space as the stimulating current, the normal component of the intracellular current density is zero, and the extracellular potential is kept constant over the electrode surface. All stimuli have a duration of 2 ms. Before stimulation the resting potential is equal to EK. In each simulation, the initial stimulus (S1) is cathodal with twice the rheobase stimulus strength. Rheobase depends on the model parameters, so the S1 strength in mA/cm varies between simulations.

All movies were generated using 1 ms per frame. Only one quadrant of the tissue is shown. The transmembrane potential in the other quadrants can be found from the even symmetry. Movies were generated using an S2 strength about 10% above the threshold strength for that interval.

Space-clamped simulations were performed using Eq 9 with the left-hand-side set to zero, and the stimulus was applied through an additional membrane current added to IK and INa.

When equal anisotropy ratios were used, the conductivities were chosen to be, gix = gex = 2 mS cm-1, and giy = gey = 0.32 mS cm-1.

The bidomain equations are solved numerically with the finite difference method. Using an initial value of Ve(t), we solve Eq 9 for Vm(t+Δt). Then we solve Eq 10 for Ve(t+Δt) using the value of Vm(t+Δt) in the source term, and the successive overrelaxation (SOR) method [4,6]. In order to accelerate convergence, an overrelaxation parameter of w = 1.8 was used [24]. The iterative loop terminates when changes in Ve between subsequent time steps is less than 1 μV. All programs are written in Fortran 90 and compiled using the gfortran compiler (www.gcc.gnu.org). The space step is 0.005 cm parallel to the fiber direction and 0.002 cm perpendicular to the fiber direction. The time step is 1 μs. The number of nodes in x-direction is 200 and the number of nodes in y-direction is 500. Because the tissue is stimulated from the center and propagation is symmetrical, we consider only one quadrant of the tissue.

During simulations of anodal excitation, we encountered a computational problem arising from the strong hyperpolarization next to the electrode. The time constant τh became too small (even smaller than the time step), making the calculations unstable (Fig 16). In order to correct this, we modified the parsimonious ionic model by requiring that the minimum value of τh be the time step (1 μs). This modification made the simulations stable.

Supporting information

S1 Movie. Cathode make excitation.

S1-S2 interval = 155 ms, S2 stimulus strength = 33 mA/cm, S2 stimulus polarity = cathodal.

doi:10.1371/journal.pone.0171144.s001

(MP4)

S2 Movie. Cathode break excitation.

S1-S2 interval = 143 ms, S2 stimulus strength = 435 mA/cm, S2 stimulus polarity = cathodal.

doi:10.1371/journal.pone.0171144.s002

(MP4)

S3 Movie. Anode make excitation.

S1-S2 interval = 158 ms, S2 stimulus strength = 640 mA/cm, S2 stimulus polarity = anodal.

doi:10.1371/journal.pone.0171144.s003

(MP4)

S4 Movie. Anode break excitation.

S1-S2 interval = 130 ms, S2 stimulus strength = 710 mA/cm, S2 stimulus polarity = anodal.

doi:10.1371/journal.pone.0171144.s004

(MP4)

S5 Movie. Failed anode break excitation.

S1-S2 interval = 125 ms, S2 stimulus strength = 600 mA/cm, S2 stimulus polarity = anodal.

doi:10.1371/journal.pone.0171144.s005

(MP4)

S6 Movie. Anode break excitation.

S1-S2 interval = 130 ms, S2 stimulus strength = 750 mA/cm, S2 stimulus polarity = anodal. kr = 21.28 mV and gK = 0.3 mS cm-2.

doi:10.1371/journal.pone.0171144.s006

(MP4)

S7 Movie. Failed anode break excitation.

S1-S2 interval = 130 ms, S2 stimulus strength = 750 mA/cm, S2 stimulus polarity = anodal. kr = ∞ and gK = 0.016 mS cm-2.

doi:10.1371/journal.pone.0171144.s007

(MP4)

S8 Movie. Anode make/break excitation for EK = -69 mV.

S1-S2 interval = 200 ms, S2 stimulus strength = 500 mA/cm, S2 stimulus polarity = anodal.

doi:10.1371/journal.pone.0171144.s008

(MP4)

Acknowledgments

This research was supported in part by a grant from the National Institutes of Health (R01HL118392).

Author Contributions

  1. Conceptualization: SKG RAG BJR.
  2. Data curation: SKG.
  3. Formal analysis: SKG.
  4. Funding acquisition: BJR.
  5. Investigation: SKG RAG BJR.
  6. Methodology: SKG RAG BJR.
  7. Project administration: BJR.
  8. Resources: BJR.
  9. Software: SKG.
  10. Supervision: BJR.
  11. Validation: SKG RAG BJR.
  12. Visualization: SKG.
  13. Writing – original draft: SKG.
  14. Writing – review & editing: SKG RAG BJR.

References

  1. 1. Fenton F, Cherry E. Models of cardiac cell. Scholarpedia. 2008;3: 1868.
  2. 2. Kandel SM, Roth BJ. The strength-interval curve in cardiac tissue. Comput Math Methods Med. 2013;2013: e134163.
  3. 3. Roth BJ. Strength-interval curves for cardiac tissue predicted using the bidomain model. J Cardiovasc Electrophysiol. 1996;7: 722–737. pmid:8856463
  4. 4. Beeler GW, Reuter H. Reconstruction of the action potential of ventricular myocardial fibers. J Physiol (Lond). 1977;268: 177–210.
  5. 5. Roth BJ. A mathematical model of make and break electrical stimulation of cardiac tissue by a unipolar anode or cathode. IEEE Trans Biomed Eng. 1995;42: 1174–1184. doi: 10.1109/10.476124. pmid:8550059
  6. 6. Wikswo JP, Lin SF, Abbas RA. Virtual electrodes in cardiac tissue: a common mechanism for anodal and cathodal stimulation. Biophys J. 1995;69: 2195–2210. doi: 10.1016/S0006-3495(95)80115-3. pmid:8599628
  7. 7. Kandel SM, Roth BJ. Intracellular calcium and the mechanism of the dip in the anodal strength-interval curve in cardiac tissue. Circ J. 2014;78: 1127–1135. pmid:24583915
  8. 8. Luo CH, Rudy Y. A dynamic model of the cardiac ventricular action potential. I. Simulations of ionic currents and concentration changes. Circ Res. 1994;74: 1071–1096. pmid:7514509
  9. 9. Joung B, Park HW, Maruyama M, Tang L, Song J, Han S, et al. Intracellular calcium and the mechanism of anodal supernormal excitability in langendorff perfused rabbit ventricles. Circ J. 2011;75, 75: 834, 834–843. pmid:21301131
  10. 10. Ranjan R, Chiamvimonvat N, Thakor NV, Tomaselli GF, Marban E. Mechanism of anode break stimulation in the heart. Biophys J. 1998;74: 1850–1863. doi: 10.1016/S0006-3495(98)77895-6. pmid:9545047
  11. 11. Ranjan R, Tomaselli GF, Marbán E. A novel mechanism of anode-break stimulation predicted by bidomain modeling. Circ Res. 1999;84: 153–156. pmid:9933246
  12. 12. Shotwell MS, Gray RA. Estimability analysis and optimal design in dynamic multi-scale models of cardiac electrophysiology. JABES. 2016;21: 261–276. doi: 10.1007/s13253-016-0244-7. pmid:27330268
  13. 13. Gray RA, Pathmanathan P. A parsimonious model of the rabbit action potential elucidates the minimal physiological requirements for alternans and spiral wave breakup. PLoS Comput. Biol. (in press)
  14. 14. Gray RA, Huelsing DJ, Aguel F, Trayanova NA. Effect of strength and timing of transmembrane current pulses on isolated ventricular myocytes. J Cardiovasc Electrophysiol. 2001;12:1129–1137. pmid:11699521
  15. 15. Nygren A, Giles WR. Mathematical stimulation of slowing of cardiac conduction velocity by elevated extracellular [K+] in a human atrial strand. Ann Biomed Eng. 2000;28:951–957. pmid:11144680
  16. 16. Sepulveda NG, Roth BJ, Wikswo JP. Current injection into a two-dimensional anisotropic bidomain. Biophys J. 1989;55: 987–999. doi: 10.1016/S0006-3495(89)82897-8. pmid:2720084
  17. 17. Trayanova N, Skouitine K. Modeling defibrillation: Effects of fiber curvature. J Electrocardiol. 1998;31(suppl 1): 23–29.
  18. 18. Galappaththige S, Roth BJ. Electrical pacing of cardiac tissue including potassium inward rectification. PLoS ONE. 2015;10: e0127837. doi: 10.1371/journal.pone.0127837. pmid:26057242
  19. 19. Dekker E. Direct current make and break thresholds for pacemaker electrodes on the canine ventricle. Circ Res. 1970;27: 811–823. pmid:5486248
  20. 20. Sidorov VY, Woods MC, Wikswo JP. Effects of elevated extracellular potassium on the xtimulation mechanism of diastolic cardiac tissue. Biophys J. 2003;84: 3470–3479. doi: 10.1016/S0006-3495(03)70067-8. pmid:12719272
  21. 21. Roth BJ, Patel SG. Effects of elevated extracellular potassium ion concentration on anodal excitation of cardiac tissue. J Cardiovasc Electrophysiol. 2003;14: 1351–1355. pmid:14678112
  22. 22. Roth BJ. Artifacts, assumptions, and ambiguity: Pitfalls in comparing experimental results to numerical simulations when studying electrical stimulation of the heart. Chaos. 2002;12: 973–981. doi: 10.1063/1.1496855. pmid:12779621
  23. 23. Henriquez CS. Simulating the electrical behavior of cardiac tissue using the bidomain model. Crit Rev Biomed Eng. 1993;21: 1–77. pmid:8365198
  24. 24. Press WH, Teukolsky SA, Betterling WT, Flannery BP. Numerical recipes in Fortran 77: the art of scientific computing. 2. ed., reprinted with corr. Cambridge: Cambridge Univ. Press; 2003.