Complex structure of electrophysiological gradients emerging during long-duration ventricular fibrillation in the canine heart

Abstract

Long-duration ventricular fibrillation (LDVF) in the globally ischemic heart is a common setting of cardiac arrest. Electrical heterogeneities during LDVF may affect outcomes of defibrillation and resuscitation. Previous studies in large mammalian hearts have investigated the role of Purkinje fibers and electrophysiological gradients between the endocardium (Endo) and epicardium (Epi). Much less is known about gradients between the right ventricle (RV) and left ventricle (LV) and within each chamber during LDVF. We studied the transmural distribution of the VF activation rate (VFR) in the RV and LV and at the junction of RV, LV, and septum (Sep) during LDVF using plunge needle electrodes in opened-chest dogs. We also used optical mapping to analyze the Epi distribution of VFR, action potential duration (APD), and diastolic interval (DI) during LDVF in the RV and LV of isolated hearts. Transmural VFR gradients developed in both the RV and LV, with a faster VFR in Endo. Concurrently, large VFR gradients developed in Epi, with the fastest VFR in the RV-Sep junction, intermediate in the RV, and slowest in the LV. Optical mapping revealed a progressively increasing VFR dispersion within both the LV and RV, with a mosaic presence of fully inexcitable areas after 4–8 min of LDVF. The transmural, interchamber, and intrachamber VFR heterogeneities were of similar magnitude. In both chambers, the inverse of VFR was highly correlated with DI, but not APD, at all time points of LDVF. We conclude that the complex VFR gradients during LDVF in the canine heart cannot be explained solely by the distribution of Purkinje fibers and are related to regional differences in the electrical depression secondary to LDVF.

ventricular fibrillation (VF) is a major cause of sudden cardiac death. The spatiotemporal organization of VF evolves as global ischemia progresses and alters electrophysiological properties of the myocardium. Concurrent with the metabolic and electrophysiological changes, the chance of successful defibrillation and resuscitation diminishes with every minute of VF, approaching zero at ∼10 min after the onset of VF (31). Unfortunately, the delayed response time of emergency medical services often approaches the time when successful resuscitation is unlikely. This perhaps explains the fact that the overall survival of victims of out-of-hospital cardiac arrest is a dismal 5% (14). The mechanisms determining successful defibrillation and survival under these conditions remain poorly understood but may be related to the nature of the sources maintaining VF (reentrant or focal) and electrical depression caused by ischemia, culminating in a complete loss of electrical response (asystole). The evolution of the organization of VF lasting 10–20 min in the globally ischemic heart [termed long-duration VF (LDVF)] has been the focus of several recent studies (2, 5, 8, 15, 19, 20, 27, 34, 42). These studies revealed large gradients in the VF activation rate (VFR) between the left ventricular (LV) endocardium (Endo) and LV epicardium (Epi) in the dog and the rabbit but not in the pig. In the rabbit heart, a transmural VFR gradient is present only in the LV but not in the right ventricle (RV) (42). Based on principal differences in the organization of Purkinje fibers in these species (exclusively Endo in the dog and rabbit vs. transmural in the porcine heart), as well as analysis of the directionality of the transmural wave propagation in the dog and pig, the differences in the organization of LDVF between the dog and pig were interpreted in terms of the leading role of focal activity (e.g., abnormal automaticity) arising from Purkinje fibers in maintaining advanced stages of LDVF (2, 8, 34). The idea of focal sources driving VF is a significant shift of the current paradigm of VF based on the notion of reentry in the form of spiral waves and may have important implications for defibrillation and resuscitation during LDVF. Recently, an elegant study (19) using a transmural multilevel optrode in the LV showed that the gradient in VFR is mostly determined by the gradient in the diastolic interval (DI), whereas the action potential (AP) duration (APD) remains conserved across the LV wall as the LDVF evolves. This can be interpreted as evidence that a gradient of excitability and/or postrepolarization refractoriness occurs across the LV wall, which enables Endo to support more rapid activations during LDVF. The presence of such a gradient may be due to the transmural distribution of ionic channel properties unrelated to but coincidental with the distribution of Purkinje fibers. Thus, the question arises as to whether the sites of the fastest activation at advanced stages of LDVF are universally associated with Endo, where Purkinje fibers are localized in the canine heart.

In that regard, it is important to note that in addition to the transmural VFR gradient, optical mapping studies (4, 42) in the rabbit heart have revealed a large Epi VFR gradient between the RV (faster activation) and LV during LDVF. However, little is known about right-to-left gradients during LDVF in large mammalian hearts, in particular in the canine heart, which is considered to be the closest to the human heart in terms of the dynamics of LDVF (2). The purpose of this study was to provide a comprehensive description of both transmural and lateral [RV to septum (Sep) to LV] electrophysiological gradients during LDVF in the canine heart during the first 10 min of LDVF using information from both transmural needle electrodes and high-resolution Epi optical mapping. From optical data, we analyzed the distribution of APD, DI, and sustained sources of activation in the RV and LV during LDVF. Our results revealed an unprecedented complexity of electrophysiological gradients emerging during LDVF, with large heterogeneities both between chambers as well as locally within each chamber. Importantly, the magnitudes of interchamber and intrachamber VFR gradients were comparable to the magnitude of the transmural gradient, with Epi sites at the RV-Sep junction being activated as fast as LV Endo. We also found that Epi right-to-left VFR gradients were determined by the gradients in DI amid relatively constant APD, which is similar to the pattern observed transmurally across the LV wall (19). We concluded that VFR gradients during LDVF in the canine heart cannot be fully explained by special properties of Purkinje fibers but may be universally determined by nonuniform electrical depression reflected in the prolongation of postrepolarization refractoriness and eventual loss of excitability. We propose possible factors contributing to the highly nonuniform response of the canine heart to the stress imposed by LDVF and global ischemia.

METHODS

This investigation conformed with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (NIH Pub. No. 85-23, Revised 1996). The animal protocols were approved by the Institutional Animal Care and Use Committee of the University of Utah. A total of 19 dogs were used in this study (5 dogs for the multielectrode mapping of LDVF in situ, 10 dogs for the optical mapping of LDVF in isolated hearts, and 4 dogs for the assessment of regional differences in refractoriness during global ischemia in isolated hearts).

Experiments in situ.

Five dogs of either sex (30.8 ± 10.5 kg) were anesthetized with acepromazine (0.1 ml/10 kg) and pentobarbital sodium (32.5 mg/kg). Anesthesia was maintained through intravenous injections of pentobarbital sodium. Intubation was accomplished with a cuffed endotracheal tube, and mechanical ventilation was maintained at 10–15 cycles/min (tidal volume: 15–20 ml/kg). The heart was exposed via a midline sternotomy. The chest opening was covered with polyethylene film, and warm, humidified air was blown into the chest to maintain temperature at the surface of the heart between 35 and 37°C during LDVF. Lead I ECG was recorded continuously throughout the experiment. Blood gases, pH, electrolyte concentrations, and glucose were analyzed in arterial blood samples before the onset of VF. Plunge needle electrodes with 10 evenly spaced unipolar leads were manufactured in house following the design developed by Rogers et al. (30). Electrodes in the LV had an interlead distance of 1.6 mm, and those in the RV and Sep had an interlead distance of 1.2 mm. Three needle electrodes were inserted in the anterior RV, one in the Sep, and three in the anterior LV, as shown schematically in Fig. 1A. The distance between needles was 10–15 mm. Needles were placed along a transverse plane bisecting the heart approximately at half-distance between the apex and base. The Sep needle was placed just to the right of the left anterior descending coronary artery (LAD). VF was induced by a brief (∼1 s) application of current from a 9-V battery to the RV outflow tract. Unipolar electrograms from all contacts of the needle electrodes were recorded continuously during the first 10 min of VF at a sampling rate of 1 kHz using a custom-made multichannel data-acquisition system, as previously described (35).

Fig. 1.

Schematic representation of the mapping modalities used in this study. A: approximate positions of the plunge needle electrodes. RV1, RV2, and RV3 indicate electrodes in the right ventricle (RV). LV1, LV2, and LV3 indicate electrodes in the left ventricle (LV). The electrode in the anterior projection of the interventricular septum (Sep) is also shown. All electrodes were in the transverse plane located at approximately middistance between the base and apex. The epicardial (Epi) and midmyocardial (Mid) leads in the plunge needle electrodes are indicated, as is the endocardial (Endo) lead in the RV and LV electrodes. Deep indicates the lead in the Sep electrode that is the most distant electrode from Epi and is located in the depth of Sep. See text for more detail. B–D: three positions of the imaged area (circles) used in the optical mapping experiments. B: the imaged area mostly covers the LV with a narrow region of the RV. C: the imaged area covers approximately equal portions of the RV and LV. D: the imaged area mostly covers the RV with a narrow region of the LV. LAD, left anterior descending coronary artery.

Electrode data analysis.

Unipolar electrograms were analyzed using custom software developed in a Matlab framework. Activations were chosen as the maximum negative derivatives that exceeded a minimum of 3 V/s within a 20-ms search window, which is similar to previously described approaches (8, 9). VFR was calculated as the average number of activations per second over 10-s intervals taken at 10 s after VF induction and at minutes 1-9 of LDVF. For simplicity, the first time point is referred to as minute 0 of LDVF.

Whereas the length of the RV and LV needles was constant and designed to span the largest thickness observed in the respective chambers of the canine heart, the actual thickness of the ventricular wall is highly nonuniform, especially in the RV. Thus, in some cases, plunge needle electrodes were longer than the wall thickness at the site of insertion. The contacts outside the ventricular wall were excluded from analysis based on the following criteria: 1) low VFR at 0 min of LDVF (slower by 5 Hz or more compared with the average VFR measured in all locations), 2) decrease of VFR to below 2 Hz within the first minute of LDVF, and 3) observation of a progressively decreasing electrogram amplitude toward more distal leads while maintaining the same morphology as more proximal leads in the same needle. In the latter case, the most proximal lead from the group of leads with identical morphology was designated as the Endo location in the needle. This approach could potentially falsely reject some of the true Endo leads in the RV and LV and thus underestimate the magnitude of the Endo-to-Epi difference in VFR, which, however, would not affect the main conclusions of this study (this issue is further discussed in Limitations). The position of the Epi lead could be easily ensured by visual inspection. After determination of the most distal lead in contact with the myocardial wall (Endo), the midmyocardial (Mid) lead was chosen as the one closest to being equidistant between Epi and Endo. Note that, unlike in RV or LV needles, the most distal recording site in the Sep needle was always in the depth of the Sep (see Supplemental Material, Supplemental Fig. 1).¹ Accordingly, in the Sep needle, we identified the Epi, Mid, and Deep sites for the purposes of analysis and comparison with RV and LV locations. At later stages of LDVF, some electrograms appeared to lack any local activation. However, due to a far-field signal always being present in unipolar electrograms, it was sometimes difficult to ascertain a clear-cut case of complete loss of excitability. Therefore, local inexcitabilty was defined in unipolar electrograms when VFR fell below 0.5 Hz, whereas at least some other leads in the same experiment exhibited fibrillatory activity with VFR > 4 Hz.

Experiments in isolated hearts.

Hearts were obtained from dogs of either sex (n = 14, 27.5 ± 2.7 kg) after premedication, anesthesia, and surgery performed as described above. After isolation via a midline sternotomy, the heart was perfused in a Langendorff apparatus with a mixture of blood and Tyrode solution as described in detail in our previous publications (16, 43). Briefly, the blood-Tyrode mixture was oxygenated (5% CO₂-95% O₂), heated (37°C), and filtered using standard pediatric heart perfusion units from various vendors. The whole mixture was collected for recirculation with collector tubes inserted into the RV and LV via cuts in the appendages of the respective atria. The heart was then placed in a temperature-controlled bath with heated water-jacketed transparent glass walls. The bath was filled with warmed Tyrode solution, which was continuously pumped, without recirculation, at a rate 80–150 ml/min. The O₂ content in the superfusate was removed by continuous gassing with a 95% N₂-5% CO₂ mixture. Temperatures in the LV cavity, superfusate, and water jacket were maintained at 37 ± 0.5°C during both normal coronary perfusion and ischemia. The gradient of temperature across the LV wall did not exceeded 1°C. Ten hearts were used for the optical mapping of LDVF. In four additional hearts, we performed programmed stimulation during global ischemia to access regional differences in refractoriness (see Supplemental Figs. 6–8).

Optical recordings.

Optical mapping of activation during VF was performed using an electron-multiplied charge-coupled device camera (iXon DU-860D, Andor Technology, Belfast, UK) with a 6- or 12-mm objective lens (Computar, Commack, NY) at a resolution of 64 × 64 pixels and frame interval of 2 ms. Excitation light came from a 532-nm green solid-state laser (Coherent, Santa Clara, CA), and the fluorescent signal was filtered with a 640 ± 50-nm filter (Omega Optical, Brattleboro, VT). The voltage-sensitive dye di-4-ANEPPS (Molecular Probes, Carlsbad, CA) was delivered to the heart as a bolus via an injection into the aortic cannula. The field of view covered portions of Epi of the anterior RV, Sep, and LV. Even though the sites immediately adjacent to the LAD most certainly overlay the Epi projection of the Sep, the exact boundaries between the Sep and RV and LV free walls could not be reliably determined for the purposes of optical mapping. Therefore, in optical maps, we followed previously used approaches (4, 32, 42) and defined the area to the right of the LAD as the RV and the area to the left of the LAD as the LV (Fig. 1B). The size of the field of view varied between 40 × 40 and 65 × 65 mm. Due to the relatively large size of the dog heart and curvature of the ventricular wall, it was not possible to have large portions of both ventricles simultaneously in the field of view. Therefore, optical mapping experiments were divided into three subgroups. In the first group (n = 4), the field of view predominantly covered the LV with a rim of the RV ∼15 mm wide. In the second group (n = 3), the areas to the right and left of the LAD were approximately equal. In the third group (n = 3), the field of view predominantly covered the RV with a rim of the LV ∼15 mm wide. The approximation of the respective fields of view is schematically shown in Fig. 1B. To minimize motion artifacts, the heart was gently pressed against the glass chamber wall as previously described (16, 43). No electromechanical uncouplers were used.

The heart was perfused with blood before the onset of LDVF; however, immediately before the induction of ischemia, the perfusion was switched to Tyrode solution with the dual purpose of ensuring constant composition of the extracellular milieu at the onset of LDVF and enhancing the voltage-sensitive signal for the purpose of better resolving low-amplitude signals at advanced stages of LDVF. VF was induced using a 9-V direct current battery 10 s to 5 min before the onset of global ischemia initiated by interruption of aortic perfusion. No perceptible differences in the measured parameters were found between experiments depending on the duration of VF before the onset of ischemia; therefore, in all experiments, the time of LDVF was counted from the moment when perfusion was stopped. LDVF was maintained for 10 min; 4–6-s-long movies were acquired every 30 s. After the last LDVF movie, the perfusion pump was restarted, and additional movies were acquired every minute during 5 min of reperfusion.

Analysis of optical signals.

Spatial distributions of VFR, APD, and DI at different time points during LDVF were estimated in optical mapping data using custom routines developed in PV-Wave software (Visual Numerics, Boulder, CO). The dominant frequency of the Fourier spectrum was not used to estimate VFR because it did not reliably represent the perceived number of activations per unit of time at advanced stages of LDVF, when the AP is characterized by very short APD and long DI. Instead, we used an algorithm to detect individual APs in single-pixel recordings, similar to that described in our previous publication (16). In brief, the depolarization and repolarization phases of each optical AP were detected. The time points at which the depolarization and repolarization phases crossed a line drawn at the 40% level of the absolute maximum in the respective signal (ΔF_max) were determined. From these time points, APD, DI, and VF cycle length (VFCL) could be calculated for each cycle. The algorithm rejected noise based on thresholds set for APD (<10 ms), VFCL (<30 ms), absolute AP amplitude (<15 levels of brightness), and percent AP amplitude with respect to the time sequence maximum (<5% of ΔF_max). VFR maps were created by dividing the number of APs by duration of the analyzed movie. Areas with no APs detected throughout the entire movie were considered to be inexcitable and having a VFR = 0. Such areas were counted for the purposes of determining the average VFR but were excluded from the calculation of average APD and DI. Singularity points were identified in Hilbert-transformed movies of voltage-sensitive fluorescence as points where all AP phases converged, as previously described (13, 39). Rotors were defined as singularity points that lasted for more than one cycle of rotation. The maximal lifespan and maximal number of rotations of rotors were computed for the RV and LV at 0 and 6 min of LDVF.

Statistical analysis.

Within the data collected from the plunge needle electrodes, the time course of VFR was statistically compared in a 3 × 3 grid of principal locations (RV Epi, RV Mid, RV Endo, Sep Epi, Sep Mid, Sep Deep, LV Epi, LV Mid, and LV Endo; see Fig. 1A). In each experiment, the RV and LV were represented by average values from all three electrodes in the respective chambers. Two-way ANOVA was performed with a post hoc Bonferroni test to make all 36 pairwise comparisons of the VFR time course between the 9 locations (see Supplemental Table 1). Note that the statistical significance indicated the differences between the means at all time points of LDVF in different locations. No conclusions regarding differences at individual time points were possible due to the prohibitively large number of required pairwise comparisons.

Table 1.

Maximal lifespan of repetitive reentrant patterns in epicardial optical maps

	Maximum Lifespan, ms	Maximum Number of Rotations
Right ventricle
0 min of LDVF	504 ± 188	4.3 ± 1.2
6 min of LDVF	906 ± 278*	4.7 ± 1.5
Left ventricle
0 min of LDVF	166.6 ± 3.0*	1.7 ± 0.6*
6 min of LDVF	No reentry	No reentry

Values are means ± SE. LDVF, long-duration ventricular fibrillation.

^*P < 0.05 vs. the right ventricle at 0 min of LDVF.

In the data collected from optical recordings, average values of VFR, APD, and DI were calculated for areas to the right (RV) and left (LV) of the LAD. Despite differences in the distribution of the field of view between the RV and LV in different experiments, the average values for the two areas were similar; therefore, data from all optical mapping experiments were combined into respective regions. Two-way ANOVA was performed with a post hoc Bonferroni test to compare the time course of VFR, APD, and DI during LDVF between the RV and LV. Similar to the statistical treatment of the electrode data, the statistical significance indicated the differences between the mean values of measured parameters at all time points of LDVF. A linear regression analysis was performed to correlate the mean APD versus the mean VFCL and the mean DI versus the mean VFCL throughout all time points analyzed during LDVF. Data are expressed as means ± SE. Differences at P < 0.05 were considered statistically significant.

RESULTS

Multielectrode mapping in situ.

Representative examples of unipolar electrograms obtained from plunge needle electrodes inserted into the RV, Sep, and LV are shown in Fig. 2. At 0 min of LDVF, VFR was relatively uniform among all locations and all layers. At 10 min of LDVF, a prominent Endo-to-Epi gradient in VFR was typically observed in both LV and RV free walls (Endo faster), whereas in the Sep, the fastest VFR was observed in Epi or Mid locations with slower VFR in the Sep Deep. After 3 min of LDVF, the sites of the fastest VFR in the RV occurred in Endo in four of five (80%) experiments; in one experiment, VFR was uniform across the RV wall. In the LV, the fastest sites were in Endo in three of five (60%) of experiments; in the other two experiments, Endo and Mid had similar VFRs, with both being faster than Epi VFR. In the Sep, the fastest sites were in Epi and/or Mid in 100% of the experiments, with Sep Deep locations being consistently slower than Sep Mid locations. In many cases, we observed what appeared to be a complete lack of activation in some electrograms. Such events occurred starting from ∼4 min of LDVF and were often local, so that adjacent needles or different contacts in the same needle still reflected some activation. Using our criterion for inexcitability (see methods), local inexcitability was most frequently observed in LV Epi and RV Epi (100% and 60% of experiments, respectively). Much less frequent were cases of inexcitability in LV Endo, RV Endo, RV Mid, LV Mid, and Sep Deep locations (20% of experiments in each case). Finally, inexcitability was never observed in Sep Epi and Sep Mid.

Fig. 2.

Unipolar electrograms recorded from different leads of plunge needle electrodes during long-duration ventricular fibrillation (LDVF) in a representative experiment. Top: 0 min of LDVF; bottom: 10 min of LDVF. Shown are recordings from the RV3 (left), Sep (middle), and LV3 (right) electrodes. See Fig. 1A for electrode locations and other definitions.

Figure 3 shows the average time course of VFR during the first 9 min of LDVF in the nine ventricular locations shown in Fig. 1A (RV Epi, RV Mid, RV Endo, Sep Epi, Sep Mid, Sep Deep, LV Epi, LV Mid, and LV Endo). Data for individual experiments are shown in Supplemental Fig. 2. In both Fig 3 and Supplemental Fig. 2, the wall types are distinguished by different colors (LV, red; Sep, green; and RV, blue) and the distance from the Epi surface is indicated by different symbols (Epi, circle; Mid, square; Endo, triangle; and Sep Deep, ×). Note that in Fig. 3 the VFR curves are presented redundantly in A–C and D–F. Figure 3, A–C, groups the curves by wall type and thus facilitates a visual analysis of transmural gradients. Figure 3, D–F, on the other hand, groups the same curves by depth and thus highlights the lateral gradients (between the LV, Sep, and RV) at each depth.

Fig. 3.

Time course of VF activation rate (VFR) in a matrix of 3 × 3 of nine principal locations (LV Endo, LV Mid, LV Epi, Sep Deep, Sep Mid, Sep Epi, RV Endo, RV Mid, and RV Epi) during 0–9 min of LDVF. Colors indicate different wall types (LV, red; Sep, green; RV, blue). Symbols indicate different distances from Epi (Endo, triangles; Mid, squares; Epi, circles; Sep Deep, ×). A–C: the nine VFR curves grouped by wall type (LV, Sep, and RV, respectively). D–F: the same nine VFR curves grouped by distance from Epi (Epi, Mid, and Endo/Deep, respectively). *Statistically significant difference between the VFR curves by two-way ANOVA. G: summary of statistically significant differences in VFR time courses between different locations.

Figure 3A shows the time course of VFR during LDVF at three transmural levels in the LV. Consistent with previous results, the difference between Endo, Mid, and Epi progressively increased over time of LDVF, with LV Endo being consistently the fastest location. Note, however, that the largest gradient occurred in the outer half of the LV wall, between Mid and Epi. Accordingly, the time course of VFR was significantly different (P < 0.05) between Endo and Epi and between Mid and Epi but not between Endo and Mid.

Figure 3B shows the time course of VFR at three transmural levels in the Sep. It is important to note that the Sep Deep location was at approximately the same distance from Epi as RV Endo but at least 4–5 mm away from any Endo surface (see Fig. 1A). Within the Sep, Epi and Mid VFR curves stayed very close together throughout the first 9 min of LDVF. After 3 min of LDVF, VFR in Sep Epi and Sep Mid stabilized at ∼7 Hz, whereas VFR in Sep Deep continued to fall, leading to a progressive separation of the Sep Deep curve from both Sep Epi and Sep Mid. This separation, however, did not reach statistical significance due to the relatively large variation between individual experiments (see Supplemental Fig. 2).

Figure 3C shows the time course of VFR at three transmural levels in the RV. RV Mid was faster than RV Epi throughout the entire course of LDVF, with the difference being statistically significant (P < 0.05). However, there were no significant differences between RV Endo and RV Mid. Thus, the transmural VFR gradient in RV was present mostly between Mid and Epi layers. Note that VFR in all RV layers reached a plateau after 3–4 min of LDVF, similar to Sep Epi and Sep Mid, but dissimilar to all LV layers and Sep Deep.

Figure 3, D–F, shows the same data as Fig. 3, A–C, but grouped by the distance from Epi. Figure 3D shows the VFR time course in the Epi layer of the LV, Sep, and RV. In all Epi locations, there was an initial decrease in VFR up to 3 min of LDVF, which was slightly faster in LV Epi and RV Epi than in Sep Epi. After that, however, VFR in RV Epi and Sep Epi stabilized (albeit at different levels), whereas VFR in LV Epi continued to fall. As a result, VFR was significantly different between all pairs of Epi locations (P < 0.05), with Sep Epi being the fastest, RV Epi intermediate, and LV Epi slowest.

Figure 3E shows the VFR time course in the Mid layers of the LV, Sep, and RV. There was very little difference in the VFR time course between these locations during the first 3 min of LDVF. After that, the curves slightly diverged, with Sep Mid being the fastest, LV Mid slowest, and RV Mid in between. In a way, this pattern repeats the divergence observed in Epi locations (see Fig. 3D) but at a much lower scale. As a result, there were no statistical differences between VFR time courses in all Mid locations.

Finally, Fig. 3F shows the VFR time course in Endo layers of the LV and RV and the Deep layer in the Sep. The LV Endo location was consistently faster than both RV Endo and Sep Deep, although LV Endo and RV Endo converged by 9 min of LDVF. The differences between LV Endo and RV Endo and between LV Endo and Sep Deep were statistically significant (P < 0.05). Figure 3G diagrammatically shows the VFR gradients in the directions tangential and perpendicular to the anterior ventricular wall. The results of all pairwise comparisons between the nine measured locations are shown in Supplemental Table 1.

The data presented above can be summarized as follows: 1) VFR is uniform over all locations at the onset of LDVF; 2) during LDVF, there are significant right-to-left gradients in Epi and Endo but not in Mid; 3) the largest transmural gradients in both the LV and RV develop between Mid and Epi, whereas the gradient between Endo and Mid is modest in the LV and practically nonexistent in the RV; and 4) overall, LV Epi exhibits the largest and Sep Epi the smallest decline in VFR during LDVF, so that the largest overall difference in VFR during LDVF was observed between Sep Epi and LV Epi. Figure 4 further highlights the special role of the Sep in the pattern of activation during LDVF, showing that in all experiments the highest Epi VFR was observed either in Sep Epi or the adjacent RV Epi location (electrode RV1). Finally, it is worth noting that throughout the entire length of LDVF studied, the VFR time course was very similar in LV Endo, Sep Mid, and Sep Epi (see Supplemental Fig. 3). Since Sep Epi and Sep Mid do not have Purkinje fibers and are not adjacent to LV Endo in canine hearts, these data indicate that Purkinje fiber distribution and function may not be the main determinants of the VFR time course during LDVF.

Fig. 4.

Epicardial distribution of VFR during LDVF in five experiments in situ (A–E). Lighter shades of grey indicate higher VFR. Note that in all experiments there is a local maximum in Epi VFR distribution either in the Sep or adjacent RV position (RV1) after 3–4 min of LDVF.

Optical mapping in isolated hearts.

The purpose of optical mapping experiments was to analyze the right-to-left and intrachamber gradients in VFR, APD, and DI during LDVF with high spatial resolution. Typical examples of VFR distribution maps and individual optical recordings during LDVF are shown in Fig. 5. In the experiment shown in Fig. 5A, the field of view predominantly covered the LV and a small portion of RV. At the onset of LDVF, the VFR distribution was relatively uniform, and there were no large-scale differences between the LV and RV. At 8 min of LDVF, the VFR distribution was markedly heterogeneous, with the fastest domain (∼4.5 Hz, green) situated in the RV and LV adjacent to the LAD. The largest part of the LV was activated at a VFR of ∼3 Hz (blue). Importantly, there was an area in the LV with an extremely low activation rate (<1 Hz, dark purple). The single-pixel recordings (Fig. 5A, sites a--c) were selected in such a way as to represent the three major frequency domains in the LV observed at 8 min of LDVF. At the onset of LDVF, the activation rate was similar between the three locations. However, at 8 min of LDVF, site a was the fastest, site c was intermediate, and in site b there was a single large-amplitude AP followed by a few small deflections, which are most likely due to electrotonus from the adjacent site c. Based on selected threshold criteria (see methods), our algorithm picked only the first (large) AP during the 4-s recording in site b compared with many more in sites a and c (asterisks in Fig. 6A). Note that site c, which exhibited a rhythmic series of suprathreshold APs, was only ∼2 mm away from site b, which was barely excitable. Site b became completely silent by 10 min of LDVF, whereas the RV-Sep portion and rightmost part of the LV were still excitable (not shown).

Fig. 5.

Examples of VFR distribution measured in Epi optical maps in isolated hearts at the early and late stages of LDVF. A: experiment with predominantly LV optical mapping. Top, VFR maps at 0 min (right) and 8 min (left) of LDVF. Bottom, single-pixel recordings from sites a–c (indicated on the respective VFR maps with circles). B: experiment with predominantly RV optical mapping. Top, VFR maps at 0 min (right) and 10 min (left) of LDVF. The arrow in the 10-min map indicates a thin inexcitable area (black, VFR = 0) separating two active areas (green, VFR∼ 6 Hz; blue, VFR ∼ 3 Hz). Bottom, the same layout as in A. C: single-pixel recordings of sites a and c from B shown with an expanded time scale. Note the fixed 2:1 phase relationship between activations in these two locations, suggesting that sites a and c have a common source of excitation even though they do not communicate within the imaged area. Note the extremely high level of VFR heterogeneity in both the RV and LV at the advanced stages of LDVF.

Fig. 6.

Right-to-left differences in the time courses of VFR (A), diastolic interval (DI; B), action potential duration (APD; C), and percentage of excitable area (D) in Epi optical maps during LDVF. *Statistically significant difference between respective curves by two-way ANOVA.

In the experiment shown in Fig. 5B, the field of view predominantly covered the RV and a small portion of LV. Similar to the preceding example, at the onset of LDVF, the VFR distribution was relatively uniform, and there were no large-scale differences between the LV and RV. At 10 min of LDVF, however, the VFR distribution was extremely heterogeneous. The highest VFR was observed in an apical part of RV (∼6 Hz, green). There were also two active areas in the more basal portions of the RV (∼3 Hz, blue), whereas the rest of the RV and the visible region of the LV had a VFR below 1 Hz (dark purple) or were completely inexcitable (black). Individual recordings from sites a-c showed similar fast activations at the onset of LDVF and striking differences at 10 min of LDVF. Specifically, site a exhibited fast and highly periodic activity, site b was completely silent, and site c was also highly periodic but was activated at half the rate of site a. Waves originating in the 6-Hz domain (green) did not propagate into the 3-Hz domain (blue) within the field of view. In fact, these two domains were separated by a thin inexcitable area (indicated by a white arrow in Fig. 6B) such that the waves from both domains converged and stopped at this area (not shown). Yet, sites a and c were phase locked in a 2:1 pattern, as shown in Fig. 5C. Indeed, activation in site c followed every other activation in site a with a fixed time delay. Thus, unless this is a perfect coincidence, which is highly unlikely, site c is electrically connected to site a, perhaps via an intramural excitable pathway.

A highly heterogeneous VFR distribution with patches of inexcitable areas in the RV and LV, as shown in Fig. 5, was typical for advanced stages of LDVF (after 4–5 min of no perfusion). The transitions between high and low VFR could be very sharp, with local VFR gradients reaching 4–5 Hz/mm (not shown). A common feature of VFR maps was the presence of the highest VFR in the RV localized near the LAD. Examples of heterogeneous VFR maps from each optical mapping experiment are shown in Supplemental Fig. 4. Consistent with the presence of areas with VFR = 0 in VFR maps, activation maps showed areas of no activation at advanced stages of LDVF (see examples in Supplemental Fig. 5D).

Figure 6 shows a quantitative analysis of the differences between the RV and LV that emerged during LDVF. Figure 6A shows that the average VFR was consistently higher in the RV than in the LV throughout 10 min of LDVF, with the exception of 0 min of LDVF. Figure 6B shows that average DI was longer in the LV than in the RV. Figure 6C shows that average APD was slightly shorter in the LV than in the RV between 2 and 8 min of LDVF. Note that whereas DI increased quickly during LDVF in both the LV and RV, APD changed little with time of LDVF in both chambers. Finally, Fig. 6D shows percentages of excitable areas in the LV and RV maps as a function of LDVF duration. Note that after 3 min of LDVF, the percentage of the excitable area in the LV progressively decreased, followed by a similar decrease in the RV after an ∼1-min delay. The differences between the time courses of all measured parameters in the RV versus LV were statistically significant using two-way ANOVA (P < 0.05).

Figure 7 shows that average VFCL (inverse of VFR) was strongly correlated with the duration of average DI in both chambers at all time points during LDVF. In contrast, VFCL was only weakly inversely correlated with the average APD. Thus, VFCL was not determined by APD dispersion but rather by dispersion in DI, which reflects the degree of postrepolarization refractoriness. It should also be noted that no functional relationship could be found between APD and the preceding DI at any stage of LDVF (not shown), which is consistent with previous reports (16, 19).

Fig. 7.

A and B: scatterplots of VF cycle length (VFCL) measured in the RV and LV at all time points of LDVF versus DI (A) and APD (B). There was a strong direct correlation between VFCL and DI (A) and a weak inverse correlation between VFCL and APD (B).

Stability of reentry and breakthrough patterns in optical maps.

Epi optical maps revealed multiple wavelets/short-living rotors in both chambers until relatively late stages of LDVF (∼5–6 min). During this period of time, Epi activation maps were extremely complex and fragmented with only rare occasions of complete reentrant patterns (see Supplemental Fig. 5). At 0 min of LDVF, the maximum duration and number of rotations of reentrant circuits were larger in the RV than in the LV (see Table 1). At 6 min of LDVF, the maximum number of rotations in the RV was not different from that at 0 min LDVF, whereas the maximum lifespan was longer, consistent with prolonged VFCL at this later stage of LDVF. No reentry was observed in the LV at 6 min of LDVF. In 7 of 10 experiments, repetitive breakthrough patterns were observed in the RV close to the LAD at least in one of the movies recorded after 4 min of LDVF (see Fig. 8D). The breakthrough patterns appeared in approximately the same location in at least 50% of activations recorded in the same movie, and transitions from a breakthrough to a reentrant pattern and vice versa in the same location were observed. Repetitive breakthrough patterns were not present in the RV at 0 min of LDVF and were not present in the LV throughout the entire duration of LDVF.

Fig. 8.

Examples of transitions between the focal and reentrant pattern in the RV at the late stages of LDVF. A: individual frames of a phase movie at 8 min of LDVF. In 1, several singularity points coexist in the mapped area; the point indicated by the white arrow is the leading source of activation in the mapped area for at least seven cycles. In 2, the reentrant source is replaced by a repetitive focal source in approximately the same location (black arrows). In 3, the focal source has reverted back to a repetitive reentrant source (white arrow) in approximately the same location. B: activation map at 10 min of LDVF showing a stable focal pattern (black arrows) in the same location as the focal and reentrant sources shown in A,1–3. This focal source activates the RV but not the LV. The LV is activated by a planar wave (white arrow), which is apparently unrelated to the source in the RV. Red, early activation; magenta, late activation. The red front in the basal RV (B, top) shows the excitation wave generated in the previous cycle that is exiting the field of view when a new focal wave emerges. C: unique example of stable Epi reentry in the RV (white arrow) at a very late stage of LDVF (16 min). D: approximate locations of repetitive breakthrough patterns observed in 7 of 10 hearts.

An example of transitions between focal and reentrant pattern in the same location at a late stage of LDVF are shown in Fig. 8, A–C. Figure 8A,1 shows a snapshot of a phase movie taken at 8 min of LDVF, which revealed a total of seven coexisting singularity points in the mapped area. The white arrowhead indicates the rotor, which was sustained for about seven rotations and was the dominant source in the area. The rotor was replaced by a breakthrough pattern in the same location (black arrows in Fig. 8A,2) which lasted for another 9–10 cycles. The breakthrough pattern then reverted back to reentry (white arrow in Fig. 8A,3). Figure 8B shows an activation map computed for the movie taken at 10 min of LDVF. On this map, the colors from red to magenta show a progression of wavefronts emanated by a stable and highly periodic focal source in the RV (black arrows in Fig. 9B). The focal source was situated approximately in the same site as the focal/reentrant source observed 2 min earlier and shown in Fig. 8A. The waves emitted by the RV source failed to cross the LAD and activate the LV. However, a portion of the LV was activated, at a much slower rate, by a repetitive planar wave (white arrow in Fig. 9B), which was apparently unrelated to the fast source in the RV.

Figure 8C shows a unique case of stable Epi reentry observed in the same experiment at 16 min of LDVF (and hence beyond the standard duration of LDVF analyzed in this study). The reentrant circuit was located in the basal RV not far from the LAD (counterclockwise arrow in Fig. 8C) and was sustained throughout the 6-s-long movie. Note that the reentrant waves did not penetrate into the LV, which was completely silent at this late stage of LDVF. Figure 8D shows the approximate locations of repetitive breakthrough patterns observed in 7 of 10 hearts.

DISCUSSION

This study provides an extensive analysis of heterogeneous electrical alterations occurring in the canine heart in response to the combined challenges of high excitation rate and ischemia during LDVF. The main finding of this study is the evidence of interchamber and intrachamber heterogeneities of a magnitude comparable to the much more studied gradients between Endo and Epi. In particular, our study revealed the presence of fully inexcitable areas interweaved with still excitable areas in both ventricles during the later stages of LDVF. Additionally, our optical mapping data complement and extend previous work by Kong et al. (19) showing that the right-to-left VFR gradient, similarly to the Endo-to-Epi VFR gradient, is largely determined by the nonuniform distribution of postrepolarization refractoriness amid a relatively invariant APD.

Gradients of the activation rate and sources of fastest activation during VF in globally ischemic hearts.

The progressive disparity of VFR between Endo and Epi of both the LV and RV (higher VFR in Endo) during VF in a nonperfused canine heart has been previously shown by several groups (5, 27, 41). It was also documented that RV Epi was faster than LV Epi after 3 min of LDVF and that LV Endo was faster than RV Endo during the first 10 min of LDVF (5). In the rabbit heart, there was an Endo-to-Epi gradient in the LV, but not in the RV, and a right-to-left VFR gradient was present in Epi (42). In contrast to dog and rabbit hearts, in the pig heart, LV Epi was slightly faster than LV Endo and LV Epi was slightly faster than RV Epi at least during the first 3 min of LDVF (see Fig. 3 in Ref. 27). From these results, one can conclude that the overall pattern of VFR distribution during LDVF is complex and species dependent. Yet, the most striking feature is the consistent presence of a large Endo-to-Epi gradient in the LV of both dog and rabbit hearts and the absence thereof (or a slight inverse gradient) in the LV of the pig heart. Based on the principally different arrangement of the Purkinje network (confined to Endo in the dog and rabbit and transmural in the pig), it was postulated that the transmural distribution of VFR during LDVF reflects the dominant role of Purkinje fibers in the maintenance of VF in the ischemic heart. Hence, in dog and rabbit ventricles, the rapid sources of VF are concentrated in Endo, whereas in the pig heart they are distributed throughout the wall thickness. More direct evidence for this hypothesis stemmed from experiments in which Endo ablation with Lugol solution eliminated the transmural VFR gradient (5, 8) or accelerated spontaneous VF termination in isolated LV preparation (8) and from the observation of an increasing incidence of focal patterns in intramural layers of the porcine heart during LDVF (20).

Our quantification of spatial VFR gradients in the canine LV and RV is consistent with a previous report (27) showing the simultaneous presence of both Endo-to-Epi and RV-to-LV gradients of VFR during LDVF. In addition, we found that patterns of activation of the RV and LV during LDVF differ in several respects. First, in the RV, there are no significant differences between Endo and Mid, and, thus, the VFR gradient is mostly present in the outer half of the RV wall. In the LV, however, there is a significant gradient between Endo and Mid as well as between Mid and Epi. Second, RV Epi maintains excitability for a longer time than LV Epi. Third, repetitive reentrant activity is present in the RV but not in the LV at late stages of LDVF. We also found that activation of the Sep (at least at its anterior junction with the RV and LV) deviates from the Purkinje paradigm for LDVF maintenance. In fact, the sites of fastest activation in the Sep were found within 5–6 mm from Epi, which is at least 4 to 5 mm away from the nearest Endo surface containing the Purkinje layer (see Supplemental Fig. 1). Sep Epi locations were consistently faster than surrounding RV Epi and LV Epi locations (see Figs. 3 and 4). Moreover, throughout 10 min of LDVF, the activation rate in Sep Epi and Sep Mid was similar to that in LV Endo (see Supplemental Fig. 3B). Thus, based on the fastest VFR criterion, it is possible that Sep Epi and Sep Mid layers may harbor sources maintaining LDVF. This possibility is confirmed by observations of sustained focal and reentrant sources in the vicinity of the RV-Sep junction in Epi optical maps (see Fig. 8). A similar preferential clustering of sustained Epi sources near the Sep has been previously observed during LDVF in the rabbit heart (42). Those authors attributed Sep Epi breakthroughs to a possibility that sub-Endo structures such as papillary muscle insertions located at or near the Sep can anchor and stabilize intramural reentrant sources, which would appear as stable Epi breakthroughs.

This explanation may hold true for the case of the canine heart, where a large papillary muscle is located on the right side of the Sep. However, relatively stable reentrant circuits are also present on Epi of the RV-Sep junction. We think it is possible that the special fiber arrangement with abrupt changes in fiber orientation in the RV-Sep junction may favor the stabilization of reentrant sources in this area, especially when combined with the very slow conduction velocity, short APD, and partial cell-to-cell uncoupling present during ischemia. The ability of the RV-Sep junction to support high-frequency sources during LDVF may also be related to increased mechanical stretch in this area (which is often thinner than the lateral RV wall and Sep) resulting from the sustained RV pressure increase occurring during LDVF (22, 26). Due to the activation of mechanosensitive channels and/or modulation of other ionic channels and transporters, stretch promotes conduction block and reentry and may contribute to the generation of ectopic triggers during ischemia (17, 25). It has also been shown that stretch increases VFR and/or the density of singularity points during normoxemic VF (6, 26) and thus potentially can modulate the organization of VF in ischemic hearts. In conclusion, the unique pattern of activation of the RV-Sep junction during LDVF may be due to a combination of structural and mechanical properties of this region that set it apart from the adjacent RV and LV free walls. However, these assumptions await experimental confirmation.

Mosaic inexcitability during LDVF.

A novel finding of this study is the highly heterogeneous pattern of electrical depression in Epi of both the RV and LV during later stages of LDVF (see Fig. 6). Indeed, even though on average VFR was higher in the RV than in the LV, in both chambers there were fully unexcitable regions interweaved with regions maintaining electrical activity. Such areas of local inexcitability were not observed during LDVF in porcine hearts (16). Clearly, these local Epi gradients cannot be explained in terms of Purkinje fiber distribution and function. The mechanisms responsible for these local gradients remain unknown. One possible mechanism is related to the heterogeneous activation of ATP-sensitive K⁺ channel current (I_K,ACh) secondary to mitochondrial depolarization during ischemia caused by opening of the mitochondrial inner membrane anion channel (IMAC) (1). During global ischemia in the guinea pig heart, an agonist of IMAC, FGIN-127, exacerbated the electrical depression caused by ischemia and induced regional inexcitability at ∼10 min of ischemia. It should be noted, however, that the study by Akar et al. (1) did not directly demonstrated the link between regional inexcitability and mitochondrial depolarization. A study (24) using confocal microscopy to image the mitochondrial potential in rat hearts revealed a mosaic presence of cells with depolarized mitochondrial potential among cells with mitochondria still fully polarized. However, the cells exhibiting mitochondrial depolarization appeared only relatively late in ischemia (∼20–30 min) and were observed much more frequently upon reperfusion. In addition, they did not form macroscopically continuous regions, which could explain the inexcitable areas observed in our experiments. Thus, the link between mitochondrial depolarization and electrical depression in early ischemia remains unconfirmed. We cannot exclude, however, that the electrical depression dependent on IMAC channel activation is more pronounced under conditions used in our study (LDVF in canine hearts) than in previous studies using small mammalian hearts without VF. It is also possible that the regional inexcitability in Epi of the canine ventricles is promoted by partial cellular uncoupling, which may help to establish a sharp separation between inexcitable and excitable areas. Canine ventricular Epi may be more susceptible to uncoupling during ischemia due to lower expression of connexin43 in Epi compared with deeper layers of the ventricular wall (29). The inexcitable areas observed in this study may hamper the ability of defibrillation shocks applied during advanced stages of LDVF to synchronize the ventricles and may also facilitate the formation of reentrant circuits, which would reinitiate VF after shock application.

Right-to-left gradients in APD and DI during LDVF.

To the best of our knowledge, this is the first reported analysis of right-to-left differences in DI and APD during LDVF in the canine heart. In general, VFCL reflects the local effective refractory period (ERP) under the assumption that the tissue is activated as soon as it is capable of conducting an electrical wave. This assumption is reasonable if conduction is not restricted by anatomic barriers and is sustained predominantly by functional reentrant circuits. With the normal availability of fast Na⁺ current (I_Na), the tissue is able to generate an AP almost immediately upon repolarization from previous excitation, ERP approximates APD well, and, therefore, DIs are short during VF. However, as ischemia develops and causes a decrease in excitability, ERP extends beyond APD (postrepolarization refractoriness). Consequently, DIs during ischemic VF are prolonged (19, 28, 43). Thus, one can assume that in the ischemic heart, the length of DI during VF is a reasonably good indicator of local postrepolarization refractoriness.

Our results clearly show that the right-to-left gradients in VFCL are determined by gradients in DI (and, hence, presumably, by postrepolarization refractoriness) amid a relatively invariant APD over both chambers and all time points of LDVF (see Fig. 8). In fact, there was a slightly longer APD in the RV versus in the LV, which is consistent with observations of a gradient of the same direction observed during pacing in the ischemic rabbit heart (23). In any case, despite the slightly longer APD in the RV than in the LV, VFCL was shorter in the RV than in the LV. Overall, the relationship between VFCL, DI, and APD was very similar to that recently reported by Kong et al. (19). These authors showed that the transmural gradient of VFCL across the LV wall was due to the differences in DI amid a relatively constant APD. Thus, an Endo-to-Epi APD gradient observed in normoxemic canine and human hearts at physiological pacing rates (3, 12) is not present during LDVF and does not influence the VFR gradient in the canine heart. While it is not practically possible to measure ERP in a rigorous manner during the fast and irregular rhythms characteristic of LDVF, our rough estimates of ERP differences between Epi and Endo locations in the LV and RV paced at a relatively fast frequency of 4 Hz during global ischemia are consistent with the presence of ERP gradients both between Endo and Epi and between the RV and LV (see Supplemental Figs. 6–8). Taken together, the arguments above support the notion that both transmural and right-to-left VFR gradients during ischemia in the canine heart are universally determined by the distribution of postrepolarization refractoriness.

LDVF mechanism: sources versus substrate.

As mentioned above, the focus of recent relevant studies was on the role of Purkinje fibers as sources of activation maintaining LDVF. The strongest evidence supporting the leading role of Purkinje fibers in the dog heart is the elimination of the transmural gradient after the ablation of Endo with Lugol solution and also earlier spontaneous termination of LDVF in isolated slab preparations (5, 8). As we argued above, the outer layers of the RV-Sep junction may be another source of fast activations apparently unrelated to the fast sources in Endo. Thus, the observation of earlier termination in ablated isolated slabs of the LV free wall (8) may not be applicable to whole hearts with an intact Sep and RV.

It should be noted, however, that the identification of the sources is only one aspect of LDVF. The picture of LDVF will remain incomplete without an understanding of the “substrate,” i.e., the response of the ventricles to fast sources of electrical activity. Regardless of whether or not Purkinje fibers are the predominant source of electrical waves, the question remains as to why the electrical depression develops faster in some regions of the ventricle than in others. The large left-to-right and intrachamber VFR gradients shown in this study are most likely unrelated to Purkinje fiber distribution. We argued above that VFR may be regarded as an important index of local electrical depression (and ensuing postrepolarization refractoriness) that appears to be highly nonuniform in the canine heart. During ischemia and VF, a number of ionic currents as well as other factors can contribute to heterogeneous electrical depression. It has been shown that I_K,ATP activation can be different between Endo and Epi myocytes (10). We could not find any evidence of differential activation of I_K,ATP in the RV versus in the LV during ischemia in the relevant literature. However, the fact that the LV simultaneously has a shorter APD and longer DIs than the RV during LDVF (see Fig. 7) would be consistent with larger I_K,ATP in the LV than in the RV. Indeed, increased time-independent outward K⁺ current in the partially depolarized myocardium is expected to accelerate repolarization, on one hand, and to prolong postrepolarization refractoriness, on the other hand. Other K⁺ currents whose activation is enhanced under conditions of ischemia and/or a high rate of excitation may also be involved, such as Na⁺-activated K⁺ current (18). Another possible mechanism of heterogeneous electrical depression during LDVF is the dispersion in the inactivation properties of I_Na. Cordeiro at al. (7) have recently shown that Epi cells have a more negative half-inactivation voltage than Endo cells. This property confers a greater sensitivity of ventricular Epi to electrical depression caused by an elevation of extracellular K⁺ and thus may contribute to the transmural VFR gradient observed during VF in the globally ischemic heart. Although right-to-left differences in the voltage dependence of I_Na inactivation were not reported, in the rabbit heart the right-to-left VFR gradient during LDVF was reproducible by elevating K⁺ in the normally oxygenated fibrillating heart (4), indicating chamber-specific differences in the sensitivity to hyperkalemia. We have already mentioned a possible role of mechanical stretch in the maintenance of fast and sustained sources in the RV-Sep area. It is also possible that the distribution of stretch contributes to the transmural gradient of VFR and excitability, since both experimental and computational studies (11, 17, 38) have indicated larger strain in Endo. In particular, right-to-left and/or transmural gradients in mechanical stretch may modulate the spatial distribution of I_K,ATP activation, since a reduction in ATP content under ischemic conditions sensitizes ATP-sensitive K⁺ channels to stretch (21, 36, 37).

Although the mechanisms discussed above seem to be the most relevant because they are enhanced under conditions presented by LDVF, a number of heterogeneously distributed ionic channels, such as those associated with transient outward K⁺ current, the rapid component of delayed rectifier K⁺ current, and inward rectifier K⁺ current (33), may be involved in nonuniform electrical depression during LDVF.

VF evolution: comparison with previous studies.

In his seminal work, Wiggers (40) described four different stages of VF in the open-chested dog: 1) undulatory or tachysystolic (1–2 s); 2) convulsive incoordination (15–40 s); 3) tremulous incoordination (2–3 min); and 4) progressive atonic incoordination, which usually starts 2–5 min after the onset of VF. This classification was based on the visual analysis of high-speed movies of cardiac contraction. In a more recent study in the same animal model, Huang et al. (15) distinguished five stages (stages i–v) during the first 10 min of VF, based on a quantitative analysis of spatiotemporal dynamics of wavefronts extracted from multielectrode Epi maps. They found a nonmonotonic evolution of activation patterns during LDVF, with a transient phase of increased organization and the incidence of reentry between 1 and 3 min after VF onset and a rapid decrease in these parameters in later phases of VF. These authors also observed a steady decrease in the number of wavefronts as LDVF progressed concomitant with a steady increase in the number of breakhtroughs and the incidence of conduction block.

To the extent that it is possible to compare results obtained with different recording and analysis techniques, our present study is consistent with previous observations in several aspects. Similar to Huang et al. (15), we observed a transitory increase in the occurrence of reentry between ∼2 and 4 min of LDVF (see Supplemental Fig. 5, B and C). This approximately corresponds to Wigger's tremulous incoordination phase, during which he noted the presence of “contraction wavescircling around in very limited areas” (40). During the next stage (atonic incoordination), Wiggers (40) noted that some areas completely lost contractility, whereas in other areas, especially those close to large vessels and to the right side of the LAD, contractions still persisted. These early observations are fully confirmed by our optical mapping results showing a mosaic distribution of excited and nonexcited areas during this stage of LDVF, with excitability more preserved in areas close to the LAD and in the RV (see Supplemental Fig. 4). Similarly, Huang et al. (15) reported a progressive decrease in the number of wavefronts in LV Epi, which approached zero by 10 min of LDVF. Thus, different studies in dogs agree in their observation of a rapid loss of excitability in LV Epi. It is worth noting, however, that in porcine and rabbit hearts, LV Epi maintains excitability throughout the first 10 min of LDVF (16, 42). This difference may be due to interspecies differences in the ionic channel distribution/function and/or differences in the severity of metabolic stress caused by combined effects of ischemia and LDVF.

Conclusion and significance.

Recent studies of LDVF have been focused on the role of Purkinje fibers in LDVF maintenance. Our data indicate that both sources of activity and substrate during LDVF in the canine heart are not exclusively determined by the relationship between Purkinje fibers and the working myocardium. The pattern of activation becomes ever more divergent during the course of LDVF, revealing simultaneous differences between Endo and Epi, RV, LV, and Sep, and within each chamber or wall type. Regardless of whether or not Purkinje fibers maintain self-sustained electrical activity at advanced stages of LDVF, for the purposes of successful defibrillation and resuscitation, it is at least as important whether or not a critical mass of the myocardium can respond in a synchronous manner to an electric shock and any stimulus after defibrillation shock. The presence of large inexcitable areas in the LV and RV may affect the outcome of defibrillation shock and raises the possibility of forming reentrant circuits around inexcitable regions after shock and reinitiation of VF. Thus, future studies concerning LDVF should be more focused on understanding the reasons why the electrical depression develops faster in some regions of the ventricle than in others and how electrical depression heterogeneities interact with the defibrillation shock.

Limitations.

In our experiments, LDVF began when perfusion was stopped. In a patient, the termination of perfusion may be more gradual. However, most likely, this difference would affect, at most, the first 1–3 min of LDVF. The main limitation of any mapping technique is incomplete coverage of the three-dimensional pattern of ventricular activation. However, the magnitude of the observed heterogeneities could only increase if we mapped larger regions. The thickness of the ventricular wall is highly nonuniform, whereas the length of the plunge needle electrodes used in both the RV and LV was fixed to accommodate the largest wall thickness in the respective chambers. Thus, in some locations, the most Endo leads were outside the ventricular wall. Despite our best efforts to identify and exclude such leads (see methods), an error in this process could have led to an underestimation of the transmural VFR gradient in the RV and LV. However, the fact that transmural gradients were still detected in both chambers and were consistent with a previous report (5) indicates that this potential problem was of limited influence. Whereas residual motion artifacts could have been present at the early stage of LDVF despite mechanical restraint, the largest interchamber and intrachamber gradients were observed after contractility was completely abolished by the ischemic process. Therefore, the presence of motion artifacts could not have overestimated the magnitude of the gradients. Other common limitations of optical mapping were discussed in our previous publication (16).

GRANTS

This work was supported by National Heart, Lung, and Blood Institute Grant 5-R01-HL-088444 and by Nora Eccles Treadwell Foundation research grant (to A. V. Zaitsev).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).