A detailed description of the computer model is provided in the Data Supplement. Computer simulations were performed to test the possible contribution of three different electrical properties in the genesis of inferolateral J-waves (or ER-pattern). Within the modeled heart three areas were defined: lateral zone, anterior zone, and septal zone (Figure 1A). Within each area we simulated the following interventions and evaluated their effects on the ECG. By reducing the sodium current conductivity (G_Na) to 12.5% of baseline condition we tested the depolarization hypothesis, whereas by increasing the transient outward potassium current conductivity (G_to) 10-fold we tested the repolarization hypothesis. As an alternative test for the depolarization hypothesis we simulated diffuse fibrosis with consequent conduction slowing by reducing the intracellular and extracellular conductivity to 12.5% of baseline condition (cellular uncoupling). The factor of reducing the G_Na to 12.5% was chosen by a stepwise reduction of G_Na to ½, ¼, ⅛, and 1/16 and selecting the largest value at which the effects of J-point elevations were present. Similarly, the increase of G_to was selected by a stepwise increase of G_to to 5, 10, and 15 ×, whereby the smallest increase to produce J-wave elevation was chosen. We also exaggerated the simulation to a 20-fold reduction of G_Na (= 5% of baseline) and a 20-fold increase of G_to in order to evaluate the increment of pathophysiology, which may occur preceding an arrhythmic event (Haïssaguerre et al., 2008). On top of the simulations that induced the most and largest J-waves or J-point elevations, we tested the effects of “systemic” sodium channels blockers by reducing G_Na outside the affected zone to 40% of baseline. The 40% was chosen as the value in which J-waves or J-points disappeared.

In the calculated 12-lead ECGs we determined the total QRS duration (first QRS onset in any lead to last QRS end in any lead), J-point amplitude, and presence of J-point elevations and J-waves. The J-point (Jp) was defined as the top of the end QRS notch or as the point where end QRS slurring started according to the consensus report (Macfarlane et al., 2015). A J-point elevation was defined as a Jp amplitude of 0.05 mV or more in an inferolateral lead (I, II, III, aVF, aVL, and V4-V6). A J-wave was a J-point elevation (notch or slur) of 0.1 mV or more. Difference ECGs were obtained by subtracting baseline ECGs from intervention ECGs.

The experimental protocol was approved by the local Animal Experiments Committee (Academic Medical Center, University of Amsterdam) and carried out in accordance with national and institutional guidelines.

Pigs (n = 2, male, 50–60 kg) were premedicated, intubated, and ventilated. The heart was excised and perfused according to Langendorff with a (circa 1:1) blood-Tyrode's mixture (pH = 7.35–7.45). The left obtuse marginal coronary artery (OM) or left anterior descending coronary artery (LAD) was separately cannulated. See Data Supplement for more details. The cannula was connected to a separate temperature-controlled perfusion system with the same blood-Tyrode's mixture and with a side branch for the infusion of flecainide (Tambocor, 3M Nederland, Zoeterwoude, The Netherlands, 6 or 60 μM). After baseline electrophysiological recordings, we administered flecainide to the OM/LAD cannula. For the OM perfusion the concentration was 60 μM and for LAD perfusion this was 6 μM. These are calculated concentrations based on the circulating volumes in the two perfusion systems. Initially we used a high concentration. This was chosen because the recirculating system limits the time during which a drug can be infused regionally without entering the main circulating system and causing conduction slowing in the entire heart (this is a characteristic of the cardiac circulation). During the wash-in of the drug temporary intermediate concentrations are present. During this wash-in phase the observations were made at a similar degree of conduction slowing. With the infusion of a high flecainide concentration spontaneous arrhythmias occurred after a couple of minutes following the measurements. Therefore, we used a lower concentration in the following experiment and we waited until a similar activation delay occurred in the myocardial regions. Data analysis was restricted to the conditions in which a similar degree of conduction slowing was present, and therefore a similar degree of sodium blockade (independent of the final concentration of flecainide).

The left atrium was paced at a cycle length of 450 ms. An 11 × 11 electrode grid (OM perfusion) or a 9 × 12 electrode grid (LAD perfusion) was fixed to the epicardial surface overlapping the entire LV (and with the 9 × 12 grid also the anterior RV) to obtain local unipolar epicardial electrograms. Supplementary Methods and Supplementary Figure S1 provide details on the electrode configurations. The Langendorff-perfused heart was submerged in a bucket filled with perfusion fluid, containing 61 electrodes to obtain pseudo-electrocardiograms (pseudo-ECGs, Figure 1B). The reference signal was the average of all electrodes.

In the pseudo-ECGs we determined the maximum QT interval and maximum QRS duration including the J-wave. Definition for Jp was similar as described above. Because the pseudo-ECG amplitudes were about twice those of real ECGs we adjusted the J-wave criteria accordingly. J-point elevations ≥0.2 mV were denoted as J-waves. J-wave onset (Jo) was defined as time of first deviation at the end of QRS complex initiating a J-wave—notch or slur—relative to QRS onset (Macfarlane et al., 2015). The inferolateral leads constituted the six columns of electrodes opposite the LV area (n = 18) and one bottom electrode (gray boxes in Figure 5). J-wave interval was defined as Jo to end of QRS. In the difference ECG—flecainide ECG minus baseline ECG—we determined the moment and amplitude of maximum peak difference (positive/negative). In each local unipolar electrogram we determined activation times (ATs) and repolarization times as before (Figure 1B; Coronel et al., 2006). Difference AT maps were calculated as flecainide AT map minus baselines AT map. Recordings with ST segment elevation or a flat T-wave were excluded from analysis of repolarization times. Signal analysis was performed offline using software (Potse et al., 2002) based on Matlab (The MathWorks, Inc., Natick, MA, USA).

Continuous variables were presented as mean ± SD if normally distributed and as median (25th–75th percentile) if not normally distributed.

Table 1 summarizes the simulation data on QRS durations, J-wave occurrence, maximum activation time, and activation delay for each zone of the various simulations. The ultimate AT in the anterior and lateral zones were later than in the septal zone.

G_Na reduction to 12.5% in the affected region caused a delay of the AP without affecting the AP morphology (Supplementary Figure S2A). Figure 2 shows 6 ECG leads at baseline and after G_Na reduction in each zone.

G_Na reduction in the lateral zone led to J-waves in the inferolateral leads (II, III, aVF, and V6) with a maximum amplitude of 0.14 mV in lead II (Table 1). The extremity leads showed J-wave notches and lead V6 a J-wave slur. G_Na reduction in the anterior zone led to a notching J-wave in lead I only, and to J-point depressions in the inferior leads (II, III, aVF). G_Na reduction in the septal zone did not induce J-waves.

To exclude a secondary role of I_to for J-wave induction in these simulations, we performed a G_Na intervention in the lateral zone in a model lacking I_to (G_to = 0 and G_Na = 12.5% of baseline). In this model, the reduction of G_Na in the lateral zone caused similar results in J-point elevations albeit with circa 0.02 mV lower amplitudes, which can be explained by the small differences in QRS morphology between the reference ECGs of the models with and without I_to (Supplementary Figure S3). When G_Na was amply reduced to 5% of baseline in this I_to lacking model, J-wave amplitudes were increased about 2-fold compared to the model with G_Na reduction to 12.5% (Supplementary Figure S4A).

A 10-fold G_to increase caused a deeper AP notch at the epicardium of the affected region, without influencing the endocardial AP notch (Supplementary Figure S2B). Figure 3 shows the ECG results of a 10-fold G_to increase in each zone. In all simulations of G_to increase J-waves were absent. There were J-point elevations, although in fewer leads, and with lower amplitudes compared to G_Na reduction (Table 1). Overall, G_to increase did prolong the QRS duration, but did not delay activation anywhere. When G_to was amply increased to 20-fold of baseline (until loss of AP dome occurred), J-point amplitudes increased about four times followed by ST-segment elevation in the inferolateral leads (Supplementary Figure S4B) compared to a 10-fold G_to increase. The ECG changes during a G_to increase did not resemble the typical inferolateral J-wave pattern that has been associated with idiopathic ventricular fibrillation (notch or slur in leads I, II, III, aVF, aVL, and V4-V6, see methods).

Cellular uncoupling (i.e., reduction of intracellular and extracellular conductivity to 12.5% of baseline) in the affected region caused a delay of the AP without affecting the AP morphology (Supplementary Figure S2C). Figure 4 shows the ECG results of cellular uncoupling in each zone. By reducing intercellular coupling in the lateral or anterior region, small J-point notching was induced but no J-waves appeared. The difference ECGs were of intermediate amplitude compared to G_Na reduction and G_to increase. Timing of the difference ECG peak was similar as with G_Na reduction and G_to increase. Cellular uncoupling in the lateral and anterior region induced activation delays of about 15 ms and caused maximum activation in the affected zone that determined the end of QRS (Table 1).

Sodium channel blockers can attenuate inferolateral J-waves (Roten et al., 2012). Therefore, we simulated a G_Na reduction in the rest of the heart on top of each intervention in the lateral zone (last column in Figures 2–4). In the 3 simulations with G_Na reduction in the rest of the heart, pre-existing J-waves or J-point elevations shrunk or disappeared, masked by the QRS complex that widened with 27, 22, and 17 ms (G_Na reduction, G_to increase, and cellular uncoupling, respectively). G_Na reduction in the rest of the heart delayed activation in all zones with latest activation occurring outside the 3 zones (i.e., 116 ms in the G_Na reduction and G_to increase interventions) or in the lateral zone (i.e., 123 ms in the cellular uncoupling intervention).

Figure 5 shows the unfolded bucket with electrodes and some examples of pseudo-ECGs at baseline and during OM flecainide infusion (60 μM). In this heart, J-waves appeared on the inferolateral pseudo-ECG leads. One J-wave was observed just outside this area, with reciprocal J-point depressions in the other leads (Figure 5: pseudo-ECG at C). At baseline J-waves were present in 4 leads. After flecainide infusion the number of J-waves in inferolateral leads increased, while in the other leads 2 of 3 J-waves disappeared (Table 2: OM perfusion). The amplitude of the difference ECG was larger and positive in inferolateral leads compared to the other leads. After flecainide infusion the QRS duration was increased by 26 ms due to arising J-waves. In the heart with LAD perfusion (Figure 5, Table 2), some J-waves were present at baseline, albeit mainly in the most superior leads. After flecainide infusion (6 μM) all J-waves disappeared in inferolateral leads and arose in 2 leads near the RV posterior wall. The amplitude of the difference ECG was negative and larger in inferolateral leads than in the other leads (Table 2: LAD perfusion). The peak of the difference ECG also occurred earlier during LAD perfusion [15 (12–27) ms] compared to OM perfusion [35 (28–41) ms].

Figure 6 demonstrates the activation maps at baseline and during flecainide infusion (60 μM) in the OM. At baseline, the OM region was activated latest (right side of the AT map), and showed the largest conduction delay after flecainide infusion. The J-wave interval overlaps this region of delayed activation (dotted surface in flecainide AT map) and even outreaches the AT map. We quantified the activation delay by selecting a column of electrodes inside and outside the OM region (white striped boxes) and summarized the data (Table 3). After flecainide infusion the delay in AT was larger in the OM region than in the rest of the LV. In the non-perfused regions no relevant changes occurred, as expected.

Figure 7 shows the activation maps at baseline and during flecainide infusion (6 μM) in the LAD region. This is an extended map with two thirds of the map overlying the LV and one third on the left overlying the RV anterior wall. At baseline, the J-wave interval overlaps that of late activation in the RV (dotted surface) and even outreaches the AT map. After 5 min of flecainide infusion, conduction was delayed in the LAD region and inferolateral J-waves were no longer present. Table 3 demonstrates that within the perfused region (lower left white striped box in Figure 7) activation at baseline was earlier in the activation sequence compared to the region outside the selectively perfused area (right upper box in Figure 7). After flecainide infusion activation was latest in the perfused region.

The mechanism of the inferolateral J-waves is a debated issue (Wellens, 2008; Hoogendijk et al., 2013). The two prevailing hypotheses for inferolateral J-waves are focused either on depolarization (Abe et al., 2010) or repolarization, similar to Brugada Syndrome (Yan and Antzelevitch, 1996). We have shown that the amplitude of the J-point elevation is largest with G_Na reduction compared to G_to increase and cellular uncoupling. Only G_Na reduction in the lateral and anterior region caused J-point elevations that met the criteria of J-waves. This observation supports the depolarization hypothesis. Cellular uncoupling induces smaller J-point elevations (no J-waves). Since activation delays due to G_Na reduction and cellular uncoupling were similar (about 15 ms), we argue that their difference in J-point elevations may be explained by the reduced tissue conductivity following cellular uncoupling. In an area of reduced tissue conductivity the current generated by the activation wave front is smaller, resulting in a smaller potential field on the torso. Alternatively, J-waves may be caused by early repolarization of the AP since the increase in G_to resulted in minor J-point elevations, which fits with an expected current flow from endocardium toward epicardium during phase one repolarization. However, although the simulated 10-fold increase of G_to in this study is very large (see also Supplementary Figure S2) it induces J-point elevations that are only half the size of those induced by G_Na reduction, and do not reach the critical J-wave level of 0.1 mV. Also, difference ECGs had smaller amplitudes. The latter hypothesis is therefore the less likely but cannot be excluded based on our data.

Furthermore, it may be argued that I_to might play a role in the J-wave induction during a reduction of G_Na. Indeed, in a model lacking I_to, the same reduction of G_Na led to smaller J-point elevations, although the influence was minor, and might be explained by deeper S-waves already present during baseline (i.e., reference ECG). Moreover, the peak amplitudes of the difference ECGs (G_Na minus reference) were 0.06 mV larger in the I_to lacking model compared to the model with I_to (Supplementary Figure S3B), whereas the timing of the peaks was similar. We therefore conclude that in the simulation of G_Na reduction, the role of I_to in the J-wave genesis is minor and may even have an opposing effect, i.e., in presence of I_to the effects of G_Na reduction on the ECG are smaller.

Further amplification of the interventions showed that both G_Na reduction to 5% and 20-fold G_to increase lead to larger J-wave amplitudes and QRS broadening. However, increasing the G_to 20 times does also induce ST-segment elevations in inferolateral leads, which do not resemble the inferolateral J-waves under consideration in this study. Moreover, reduction of G_Na to 5% of baseline lead to typical enhancement of the J-wave notching/slurring as seen in patients during the interval preceding an event of ventricular fibrillation (Haïssaguerre et al., 2008).

The amplitude of the difference ECG is largest with G_Na reduction followed by cellular uncoupling and then G_to increase. Also, the difference ECG was larger with intervention in the anterior region than with intervention in the lateral region, and was smallest with intervention in the septal region. However, the amplitude of the difference ECG is not directly reflected in the amplitude of a J-point elevation. This implies that the timing and spread of the difference ECG contributes importantly to the change in QRS morphology. In the pig experiments, the peak of the difference ECG occurred earlier during LAD perfusion compared to OM perfusion, whereas in the computational results the peak of the difference ECG occurred earlier in the septal compared to the lateral or anterior zone. The separate perfusions in the ex vivo hearts, however, did cover slightly different regions within the heart, i.e., with LAD perfusion we influenced the anterior as well as the septal regions. Therefore, combining the computational results from the septal and anterior zones likely represents the ex vivo observations more closely. Enhanced G_to caused more QRS widening compared to a decreased G_Na or cellular uncoupling (Table 1), whereas no activation delay occurred. This indicates that early repolarization does contribute to QRS duration.

Our experimental and simulation data suggest that conduction delay resulting from reduced sodium current in the region that is activated intrinsically late in the activation sequence at baseline, i.e., the lateral region, predominantly contributes to the genesis of inferolateral J-waves, whereas conduction delay in earlier activating regions, i.e., the anterior and/or septal region, shows minor, or no J-point elevations and even J-point depression. We observed that although the AT delay in LAD perfusion was smaller than in OM perfusion it led to J-point depression rather than J-point elevations. We suggest that this difference in J-point deviation is related to the position of the region with delayed activation with respect to the anterolateral recording leads. The septal involvement causes the voltage vector during the J wave to point rightward, away from the anterolateral leads. It is similar to the explanation of how a right bundle branch block results in broad S-wave in V6. The timing of the activation within the QRS complex (normally late activation in the OM region, and normally early activation in the anterior/septal region) causes the LAD region to have a relatively small influence on the J-wave. Recently, it has been shown in a case report that J-waves in the inferior ECG leads can result from delayed activation in the basal inferior LV region, without shortening of the AP duration (Park et al., 2014). This confirms that the region of abnormality is important.

Remarkably, all different interventions—G_Na reduction, G_to increase, and cellular uncoupling—within the lateral region show similar results in polarity of J-point elevations and difference ECGs, although only G_Na reduction led to J-waves. All interventions in the anterior region led to negative difference ECGs in the inferior leads and J-point elevation (no J-waves) were only present in lateral ECG leads. The reason that the effects of interventions are smaller in the anterior and septal zones—despite similar amplitudes of difference ECGs among zones per intervention—is therefore related to the extent of masking by the QRS complex.

J-wave amplitude is modulated by heart rate (Aizawa et al., 2012), autonomic tone (Abe et al., 2010), and drugs (Haïssaguerre et al., 2009; Roten et al., 2012). It has been demonstrated in patients that sodium channel blockers like ajmaline, pilsicainide, and flecainide attenuate J-waves in the inferolateral leads (Kawata et al., 2012; Roten et al., 2012; Nakagawa et al., 2014) and broaden the QRS (Roten et al., 2012). Our simulation data explain these findings by demonstrating that G_Na reduction in the rest of the heart widens the QRS complex and consequently masks the pre-existing J-waves.

In patients with idiopathic ventricular fibrillation it has been shown that inferolateral J-waves augment after a pause and diminish at higher heart rates (Haïssaguerre et al., 2008; Nakagawa et al., 2014). It has been supposed that the attenuation of J-waves at increasing heart rate results from reduced transient outward potassium current due to the relatively slow recovery from inactivation (Koncz et al., 2014). We did not study the effect of heart rate increase, but we surmise that conduction may also play a role in this phenomenon. Shorter coupling intervals result in lower upstroke velocities of the action potentials, especially in diseased hearts (Kodama et al., 1984). It may therefore have a similar effect as administration of ajmaline and we would expect a concomitant QRS prolongation.

Nowadays, computational techniques increase in maturity and reliability and are therefore more and more powerful to enlarge our insight in physiology and pathophysiology. As a consequence, it may reduce the number of required animal experiments. However, results from computational models are co-determined by the choice of parameter settings in the model which, due to uncertainty in the experimental data underlying the model, are partly based on assumptions. Due to the assumptions, results from the computational model could not directly be translated to the clinical situation. The results of our additional experiments, however, invigorate the computational findings.

The model did account for interaction between different currents. However, we did not study combinations of factors (G_Na reduction, G_to increase, and cellular uncoupling), although these together may lead to more exacerbated J-waves, as long as they occur in the same region. Also, it has been shown that different balances of ionic current densities generate viable action potentials (Britton et al., 2013), and therefore it would be of interest to study how the effect of each factor used in this study may change within the setting of different balances of other ion current densities. Yet, our model provides valuable insight into the potential mechanism underlying inferolateral J-waves. Another study limitation is that the pig model lacks I_to and therefore it was not suitable to test the role of I_to. With this model we were, nevertheless, able to induce J-waves, indicating that I_to is not required to induce J-waves.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.