Atrial Ionic Remodeling Induced by Atrial Tachycardia in the Presence of Congestive Heart Failure
Background— Atrial fibrillation (AF) and congestive heart failure (CHF) produce discrete forms of atrial ionic remodeling. The in vivo effects of atrial tachycardia (AT) remodeling are altered by CHF. This study evaluated underlying mechanisms at the level of ionic remodeling.
Methods and Results— We studied 4 groups of dogs: (1) unpaced controls (CTLs); (2) CHF caused by 2-week ventricular tachypacing (VTP, 240 bpm); (3) AT (400 bpm ×7 days); and (4) CHF+AT (2-week VTP with AT for the last 7 days). CHF and CHF+AT groups equally increased left atrial pressure. AF duration was increased in all paced groups. Effective refractory period (ERP) was decreased by 42% in AT versus CTL but by only 24% in AT+CHF versus CHF. CHF reduced L-type Ca2+ (ICa), transient-outward (Ito), and the slow delayed-rectifier (IKs) currents while increasing the Na+-Ca2+ exchanger (INCX) and not affecting the inward-rectifier (IK1) current. AT reduced Ito and ICa while increasing IK1 and leaving IKs unaltered. The addition of AT to CHF failed to alter Ito, IKs, or INCX beyond the effect of CHF alone, decreased ICa slightly compared with CHF alone, but had smaller effects on ICa and IK1 compared with AT alone. Thus, CHF+AT, as would occur in a CHF patient who develops AF, produced an ionic remodeling pattern different from that of CHF or AT alone and from what would have been predicted from additive effects of CHF and AT.
Conclusions— The presence of CHF alters AT-induced ionic remodeling. Thus, the ionic remodeling caused by cardiac arrhythmias in the presence of cardiac pathology is not necessarily predictable from the effects of either alone, with important potential implications for understanding the pathophysiology of arrhythmias in the diseased heart.
Received December 1, 2003; de novo received March 22, 2004; accepted April 26, 2004.
Atrial tachycardia (AT)–induced electrical remodeling plays a key role in the pathophysiology of atrial fibrillation (AF).1 Alterations in ion-channel function (“ionic remodeling”), including decreased transient-outward (Ito2–4) and L-type Ca2+ (ICa2,4,5) currents and an increased inward-rectifier (IK1) current,3,4,6 contribute to AT-induced electrical remodeling and the self-perpetuating nature of AF.1 Congestive heart failure (CHF) is an important cause of AF and also causes ionic remodeling.7 Like AT, CHF also decreases Ito; however, CHF-induced remodeling differs from AT remodeling in that CHF reduces ICa to a lesser extent, decreases the slow delayed-rectifier (IKs), and increases the Na+-Ca2+ exchanger (INCX).7 Consequently, instead of decreasing action potential duration, CHF leaves atrial action potential duration unchanged or increased.7 Whereas AT-induced ionic remodeling promotes atrial reentry by decreasing the wavelength in a spatially heterogeneous fashion,1 CHF-induced ionic remodeling increases the wavelength and may contribute to AF promotion by inducing afterdepolarization-related ectopic activity due to enhanced INCX.7,8
When AF occurs in a patient with CHF, AT remodeling develops on the background of an atrial substrate conditioned by preexisting CHF. The effective refractory period (ERP) changes caused by AT in the presence of CHF are less than those in its absence.9 These observations suggest that AT-induced ionic remodeling in the context of CHF may be different from AT-induced changes in the undiseased heart; however, there are no studies in the literature comparing AT-induced ionic remodeling in subjects with CHF with AT remodeling in normal hearts. We undertook the present studies to compare ionic changes produced by AT and CHF alone with those caused by AT in the presence of CHF.
The animal model was prepared as previously described in detail.9 All animal care procedures followed Canadian Council on Animal Care guidelines and were approved by the Animal Research Ethics Committee of the Montreal Heart Institute. A total of 100 adult mongrel dogs (19 to 39 kg) were divided into control (CTL; n=27), CHF-only (n=20), AT-only (n=25), and CHF+AT (n=28) groups. Some control animals (n=3) were shams handled like CHF dogs but without pacemaker activation. Results were the same for sham and acute controls; therefore, all were grouped for analysis. Right ventricular (RV) and right atrial (RA) unipolar pacing leads were connected to pacemakers implanted in the neck, and pacing begun 24 hours later. In CHF-only dogs, the ventricular pacemaker was programmed to capture at 240 bpm for 2 weeks. In AT-only dogs, atrioventricular block was created by radiofrequency ablation, the RV was paced at 80 bpm, and the RA was paced at 400 bpm for 1 week. CHF+AT dogs were RV-paced like CHF-only dogs, but RA tachypacing (400 bpm) was added for 1 week after 1 week of ventricular tachypacing alone.
On study days, dogs were anesthetized (morphine 2 mg/kg SC; α-chloralose 120 mg/kg IV load, 29.25 mg · kg−1 · h−1) and mechanically-ventilated. ERP was measured at the left atrial (LA) appendage with 15 basic (S1) stimuli, followed by premature (S2) stimuli with 5-ms decrements (ERP=longest S1S2 failing to capture; all stimuli twice-threshold current, 2 ms). AF (defined as an irregular atrial rhythm >400 bpm) was induced by burst pacing (10 Hz, 2 ms, 4× threshold-current pulses). Mean AF duration was determined in each dog on the basis of 10 inductions for AF <5 minutes and 5 inductions for 5- to 30-minute AF. AF >30 minutes was considered sustained: cardioversion was not performed, and electrophysiological assessment was terminated. Hemodynamic data were obtained with fluid-filled catheters and disposable transducers. After in vivo assessment, hearts were excised and atrial cells were isolated from the LA free wall by circumflex-artery perfusion.7
Currents were recorded with whole-cell patch-clamp (36±0.5°C). Borosilicate-glass electrodes had tip resistances of 1.5 to 3.0 MΩ. Compensated series resistance and capacitive time constants averaged 3.1±0.9 MΩ and 263±5 ms, respectively. Cell capacitance averaged 87±2 pF (CTL), 112±6 pF (CHF), 113±4 pF (CHF+AT), and 89±3 pF (AT; n=100/group, P<0.01 for CHF and CHF+AT versus CTL and AT). To control for cell-size variability, currents are expressed as densities (pA/pF). Junction potentials averaged 9.7±0.7 mV and were not compensated. The amplitudes of time-dependent currents (eg, ICa, Ito, and INCX) were measured from peak to steady-state values after current decay.
The solution for cell-storage contained (in mmol/L) KCl 20, KH2PO4 10, dextrose 10, mannitol 40, l-glutamic acid 70, β-hydroxybutyric acid 10, taurine 20, EGTA 10, and 0.1% bovine serum albumin (pH 7.3, KOH). Tyrode’s solution contained (in mmol/L) NaCl 136, KCl 5.4, MgCl2 1, CaCl2 1, NaH2PO4 0.33, HEPES 5, and dextrose 10 (pH 7.35, NaOH). For delayed-rectifier current recording, nifedipine (5 μmol/L), 4-aminopyridine (2 mmol/L), and atropine (200 nmol/L) were added to suppress ICa, Ito, and 4-aminopyridine–dependent, muscarinic K+ currents, respectively. Dofetilide (1 μmol/L) was added for IKs recording. For Ito and IK1 recording, nifedipine was replaced by CdCl2 (200 μmol/L). Ito was studied in the presence of 10 mmol/L tetraethylammonium to inhibit the ultrarapid delayed-rectifier current. IK1 was recorded as the 1 mmol/L barium–sensitive current. Na+ current contamination was avoided by using a holding potential of −50 mV or substituting equimolar Tris-HCl for NaCl. The internal solution for K+-current recording contained (in mmol/L) potassium aspartate 110, KCl 20, MgCl2 1, MgATP 5, LiGTP 0.1, HEPES 10, sodium phosphocreatine 5, and EGTA 5.0 (pH 7.3, KOH).
The external solution for ICa recording contained (in mmol/L) tetraethylammonium chloride 136, CsCl 5.4, CaCl2 2, MgCl2 0.8, HEPES 10, and dextrose 10 (pH 7.4, CsOH). Niflumic acid (50 μmol/L) was added to inhibit the Ca2+-dependent Cl− current.10 The internal solution for ICa recording was (in mmol/L) CsCl 120, tetraethylammonium chloride 20, MgCl2 1, MgATP 5, LiGTP 0.1, EGTA 10, and HEPES 10 (pH 7.3, CsOH). INCX was recorded as previously described,7,11 with an extracellular solution containing (in mmol/L) NaCl 140, CaCl2 2, MgCl2 1, dextrose 10, HEPES 10, niflumic acid 0.1 μmol/L, and nifedipine 5 μmol/L and an internal solution containing (in mmol/L) CsCl 130, NaCl 5, MgATP 4, and HEPES 10 (pH 7.3, CsOH).
Nonlinear curve fitting was performed with Clampfit in pCLAMP6. Group data are presented as mean±SEM. Statistical comparisons were by multiway ANOVA with F tests for interaction, and t tests with Bonferroni correction were used to compare differences between individual group means. Because of the large number of comparisons possible, we show in the figures only the statistical significance of overall group differences from CTL. A 2-tailed P<0.05 indicated statistical significance.
All dogs in CHF and CHF+AT groups showed pulmonary congestion and pericardial effusion during open-chest study. Ventricular and arterial systolic pressures were reduced in CHF dogs, and LV end-diastolic, LA, and RA pressures were significantly increased (Table 1). AT dogs were hemodynamically indistinguishable from CTLs.
In Vivo Electrophysiology
AF duration was significantly prolonged in CHF, AT, and CHF+AT groups (Figure 1A). CHF increased atrial ERP at all basic cycle lengths (BCLs) (Figure 1B). AT decreased atrial ERP at all BCLs and suppressed ERP rate adaptation. AT in the presence of CHF decreased ERP by ≈24% versus CHF alone. In animals without CHF, AT decreased ERP ≈42% from CTL.
Typical Ito was recorded from CTL cells (Figure 2A). Currents were reduced in CHF (Figure 2B), AT (Figure 2C), and CHF+AT (Figure 2D) cells. Mean Ito density was significantly and similarly reduced in all 3 groups (Figure 2E). Ito activation voltage dependence was assessed from data obtained as shown in Figure 2A through 2E, based on the relation Iv=Imax(V−Vr)(Gv/Gmax), where Iv and Gv are current and conductance, respectively, at voltage V; Imax and Gmax are maximum current and conductance, respectively; and Vr is the reversal potential (Figure 2F). Vr, as evaluated by tail currents after 2.2-ms depolarizations to +50 mV, averaged −71.5±1.4 mV, −70.6±3.0 mV, −70.2±2.8, and −69.5±1.2 mV (without junction-potential correction) in CTL, CHF, CHF+AT, and AT groups, respectively (6 cells per group, P=NS). Activation V1/2 based on Boltzmann fits to data in each experiment averaged 10.8±2.7 mV (CTL), 10.3±2.0 mV (CHF), 9.9±1.3 mV (CHF+AT), and 10.3±0.9 mV (AT; 10 cells per group, P=NS). Inactivation voltage dependence was studied with 1000-ms prepulses from −70 mV, followed by 200-ms test pulses to +50 mV. Boltzmann fitting provided inactivation V1/2s averaging −29.3±1.5 mV (CTL), −29.7±2.5 mV (CHF), −30.2±1.5 mV (CHF+AT), and −29.3±0.4 mV (AT; P=NS). Ito decay time constants showed no differences among groups (Figure 2G). Time to peak current, an index of activation speed, was not changed by CHF, CHF+AT, or AT. A paired-pulse protocol with identical 150-ms depolarizations (P1, P2) from −70 to +50 mV with varying P1-P2 intervals was used to analyze recovery kinetics. Current during P2 normalized to current during P1 was a monoexponential function of P1-P2 interval (Figure 2H). Recovery time constants averaged 31.1±2.4 ms (CTL), 33.0±2.7 ms (CHF), 30.3±0.9 ms (CHF+AT), and 29.7±1.2 ms (AT; n=7 cells per group, P=NS).
Typical IKs recordings are shown in Figure 3A through 3D. CHF cells and CHF+AT cells had reduced tail (Figure 3E) and step (Figure 3F) current densities compared with CTL or AT. Voltage dependence of IKs activation (tail-current analysis) was not altered by CHF, CHF+AT, or AT (V1/2 23.1±1.7, 20.1±1.0, 20.7±1.9, and 23.4±1.3 mV, respectively, n=12 cells per group). IKs activation kinetics at +40 mV were biexponential, with slow-phase time constants averaging 2224±340, 2232±278, 1980±211, and 1979±332 ms in CTL, CHF, CHF+AT, and AT groups, respectively (P=NS, n=10 cells per group). Fast-phase time constants averaged 248±25, 243±27, 241±20, and 221±20 ms in CTL, CHF, CHF+AT, and AT dogs, respectively (P=NS, n=10 cells per group).
Representative 1 mmol/L Ba2+–sensitive IK1 recordings are illustrated in Figure 4A through 4D. IK1 density was similar in CTL and CHF, but IK1 density was increased in AT dogs (Figure 4E). In the presence of CHF (CHF+AT), AT increased IK1 less than with AT alone (P=NS versus CTL).
ICa recordings are illustrated in Figure 5A through 5D. ICa density decreased significantly with all interventions (Figure 5E), but the decrease in CHF was less than in AT, with CHF+AT having intermediate values. The voltage dependencies of ICa activation and inactivation were unaffected by CHF, CHF+AT, or AT (Figure 5F). Vr determined by extrapolation of the ascending I-V curve to the voltage axis was similar in CTL, CHF, CHF+AT, and AT dogs (62.5±1.0, 62.1±1.6, 63.5±1.4, and 61.0±0.8 mV, respectively, n=10 cells per group). Inactivation V1/2 averaged −37.3±1.3, −37.7±1.5, −39.9±0.7, and −38.9±1.2 mV in CTL, CHF, CHF+AT, and AT groups, respectively (n=8 cells per group, P=NS). Activation V1/2 averaged −4.0±0.7, −5.1±1.2, −4.3±0.9, and −5.3±0.6 mV in CTL, CHF, CHF+AT, and AT groups, respectively (n=10 cells per group, P=NS; Figure 5F). ICa recovery time constants were not different among groups: 32.6±3.5, 33.1±4.2, 32.2±1.8, and 34.0±1.0 ms (n=6 cells per group, P=NS; Figure 5G). ICa inactivation time constants were comparable (Figure 5H).
INCX recordings are illustrated in Figure 6A through 6D. After a 5-ms depolarization from −70 mV to +20 mV, INCX was recorded as an Li+-sensitive inward current during 100-ms repolarizing pulses. INCX density was increased in CHF and CHF+AT (Figure 6E). In AT, INCX density was also increased, but to a lesser and statistically nonsignificant extent.
Interactions Between CHF and AT Effects
AT-induced percentage decreases in mean current densities from CTL are indicated in Table 2. Also shown are percentage changes produced by AT in the presence of CHF, as obtained by comparing the value in CHF+AT with the value in AT only. If there were no interaction, the percentage change caused by AT should be the same for control and CHF backgrounds. In fact, the effects of AT are different from this expectation: effects in the presence of CHF are generally smaller, and in some cases (eg, for Ito and INCX), no perceptible effect is observed.
The ionic-remodeling patterns produced by AT only, CHF only, and CHF+AT relative to CTL are shown in Table 3, along with changes expected from simple additive effects: predicted fractional current relative to CTL in CHF+AT=measured fractional mean current in presence of AT only (IAT/ICTL)×measured fractional mean current in the presence of CHF only (ICHF/ICTL). These predicted changes are compared with measured fractional mean currents in CHF+AT ([FCHF+AT=ICHF+AT /ICTL], where ICHF+AT and ICTL are currents in the presence of CHF+AT and CTL, respectively), all expressed as percentage changes from CTL. Probability values for the CHF-AT interaction are shown in the last column.
The CHF+AT profile differs from CHF only by an increase in IK1 not seen with CHF alone, along with a larger decrease in ICa than with CHF only. The CHF+AT profile differs from that of AT only on the basis of a decrease in IKs not seen with AT only, smaller changes in ICa and IK1, and a larger change in INCX than with AT only. CHF+AT differs from expectations on the basis of purely additive effects by virtue of smaller than expected changes in all currents except for IKs.
Disease-related ionic remodeling is an important contributor to a variety of cardiac arrhythmias. However, cardiac arrhythmias themselves cause ionic remodeling,2,12,13 which then superimposes on underlying disease-related remodeling. We were unable to identify studies in the literature of the ionic effects of such a combined remodeling paradigm. Therefore, we examined the ionic remodeling induced by AT alone, by CHF alone and by AT+CHF, as would occur when a patient with CHF develops AF. Our results indicate that AT effects on ion-current density are altered in the presence of CHF, producing a different ionic-current profile for the combination than anticipated based on additive ionic-remodeling effects of AT and CHF alone.
Combined Remodeling Paradigms
AF frequently occurs in the presence of preexisting heart disease, so combined remodeling paradigms involving AF-induced changes superimposed on disease-related ionic remodeling are common. Shinagawa et al9 showed that AT-induced ERP alterations in the presence of CHF are smaller than in normal hearts. The present study suggests that smaller AT-induced electrophysiological effects in CHF compared with normal hearts are due to reduced ionic remodeling (Table 2).
It would be interesting to understand the mechanisms by which CHF alters AT-induced remodeling. If the same signal transduction systems were used by both CHF and AT and were near-maximally engaged by either, the response to simultaneous stimulation would be limited. Little is known about the signal transduction involved in ion-channel remodeling. A recent study suggests that a cAMP-binding response element is involved in Ito downregulation associated with cardiac memory.14 The interaction between CHF and AT-induced remodeling appears to depend on the CHF state, rather than the AF substrate per se, based on observations during recovery from CHF. After full hemodynamic recovery from experimental CHF, the AF substrate and associated interstitial fibrosis remain,15 but CHF-induced ionic remodeling disappears16 and AT-induced ERP changes become indistinguishable from those in the normal heart.15
Relation to Previous Studies and Potential Significance
The ionic remodeling we observed in response to AT only and to CHF only is compatible with previous reports. As in previous studies,2–5 we found Ito and ICa to be reduced and IKs to be unaltered by AT. We also noted increased IK1, as have previous investigators,3,4,6 but unlike our previous detailed analysis of AT-induced ionic remodeling.2 This discrepancy may be partly due to rejection in the latter study of cells with depolarized resting potentials,2 preventing inclusion of cells with very small IK1. As noted previously,7 we found that CHF decreased atrial Ito, IKs, and ICa and increased INCX.
The prevalence of AF in various CHF series ranges from 10% to 50%.17 The combination of AF and CHF may carry a worse prognosis than either alone.18 To optimize treatment for AF in the presence of CHF, it is important to understand the underlying ionic substrate. The present study indicates that ionic remodeling caused by AT in the presence of CHF is different from that caused by AT in the normal heart, with potential implications for the effects of ion channel–modulating antiarrhythmic drugs and for preventive approaches. In addition, our findings indicate the need to evaluate interactions between other forms of cardiac ionic remodeling that may occur concomitantly in humans and to understand the underlying molecular interactions.
We selected ionic currents for study (Ito, IKs, IK1, ICa,L, and INCX) based on the fact that they have been reported to be affected by CHF and/or AT remodeling.2–7 We cannot exclude the possibility that other currents not affected by either CHF or AT alone (eg, IKr, the Ca2+-dependent Cl− current, the canine ultrarapid delayed-rectifier IKur.d, or T-type Ca2+ current) could be affected by the combination of CHF+AT. We examined cells from only 1 atrial location (the LA free wall). Because of regional heterogeneity in atrial ionic electrophysiology19,20 and in the interactions between AT and CHF,9 our findings are not necessarily applicable to all atrial regions.
We induced remodeling with regular, rapid, atrial tachypacing, which does not reproduce the irregularity of atrial rhythm in AF. Although AT is believed to be the principle stimulus to AF-induced remodeling,1 direct extrapolation to AF-induced ionic remodeling should be cautious. We used specific voltage protocols and a variety of pharmacological agents to isolate currents of interest. These approaches are required for the analysis of individual currents in native cardiomyocytes and are standard in such investigations but do have the potential to affect the currents studied. Although currents should have been affected similarly for all experimental groups, this limitation must be considered in evaluating our results.
We did not assess directly the mechanisms underlying changes in ion-channel function. Although there is evidence for transcriptional downregulation as a mechanism for AT-induced decreases in ion-current density in animal models21 and in humans,3,22 Schotten et al23 did not find ICa α1c-protein to be reduced in atrial samples from AF patients. Therefore, other mechanisms at the regulatory level, such as enhanced phosphatase activity (Christ et al, unpublished observations) and alterations in Ca2+-calmodulin kinase II,24 may also be involved. Further work on mechanisms underlying AT- and CHF-induced ion-channel dysfunction would be of great interest.
We studied the CHF-AT interaction by superimposing AT on CHF. It would be interesting to reverse the order of experimentation and assess ionic remodeling that occurs when AT is applied first, followed by CHF.
We found that AT-induced ionic remodeling is altered in the presence of CHF. Therefore, ion-channel function changes caused by an intervention in the normal heart cannot necessarily be directly extrapolated to the diseased heart, with potentially important implications for understanding arrhythmia mechanisms and therapy.
This study was supported by the Canadian Institutes of Health Research, the Quebec Heart Foundation, the Mathematics of Information Technology and Complex Systems (MITACS) Network, and a Merck Pharmacology Fellowship. Dr Cha received a fellowship grant from Kosin University, Busan, South Korea. The authors thank Evelyn Landry, Nathalie L’Heureux, and Chantal Maltais for technical assistance and France Thériault for secretarial help.
Yue L, Feng J, Gaspo R, et al. Ionic remodeling underlying action potential changes in a canine model of atrial fibrillation. Circ Res. 1997; 81: 512–525.
Van Wagoner DR, Pond AL, McCarthy PM. Outward K+ current densities and Kv1.5 expression are reduced in chronic human atrial fibrillation. Circ Res. 1997; 80: 772–781.
Bosch RF, Zeng X, Grammer JB. Ionic mechanism of electrical remodeling in human atrial fibrillation. Cardiovasc Res. 1999; 44: 121–131.
Van Wagoner DR, Pond AL, Lamorgese M, et al. Atrial L-type Ca2+ currents and human atrial fibrillation. Circ Res. 1999; 85: 428–436.
Dobrev D, Graf E, Wettwer E, et al. Molecular basis of downregulation of G-protein-coupled inward rectifying K+ current (IK,ACh) in chronic human atrial fibrillation: decrease in GIRK4 mRNA correlates with reduced IK,ACh and muscarinic receptor-mediated shortening of action potentials. Circulation. 2001; 104: 2551–2557.
Li D, Melnyk P, Feng J, et al. The effects of experimental heart failure on atrial cellular and ionic electrophysiology. Circulation. 2000; 101: 2631–2638.
Shinagawa K, Li D, Leung TK, et al. Consequences of atrial tachycardia-induced remodeling depend on the preexisting atrial substance. Circulation. 2002; 105: 251–257.
Schlotthauer K, Bers DM. Sarcoplasmic reticulum Ca++ release causes myocytes depolarization: underlying mechanism and threshold for triggered action potentials. Circ Res. 2000; 87: 774–780.
Zygmunt AC, Goodrow RJ, Antzelevitch C. INaCa contributes to electrical heterogeneity within the canine ventricle. Am J Physiol (Heart Circ Physiol). 2000; 278: H1671–H1678.
Yu H, McKinnon D, Dixon JE, et al. Transient outward current, Ito1, is altered in cardiac memory. Circulation. 1999; 99: 1898–1905.
Patberg KW, Plotnikov AN, Quamina A, et al. Cardiac memory is associated with decreased levels of the transcriptional factor CREB modulated by angiotensin II and calcium. Circ Res. 2003; 93: 472–478.
Shinagawa K, Shi Y-F, Tardif JC, et al. Dynamic nature of the atrial fibrillation substrate during the development and reversal of heart failure in dogs. Circulation. 2002; 105: 2672–2678.
Cha TJ, Ehrlich JR, Zhang L, et al. Dissociation between ionic remodeling and ability to sustain atrial fibrillation during recovery from experimental congestive heart failure. Circulation. 2004; 109: 412–418.
Wang TJ, Larson MG, Levy D, et al. Temporal relations of atrial fibrillation and congestive heart failure and their joint influence of mortality: the Framingham heart study. Circulation. 2003; 107: 2920–2925.
Feng J, Yue L, Wang Z, et al. Ionic mechanisms of regional action potential heterogeneity in the canine right atrium. Circ Res. 1998; 83: 541–551.
Li D, Zhang L, Kneller J, et al. Ionic mechanism of repolarization differences between canine right and left atrium. Circ Res. 2001; 88: 1168–1175.
Yue L, Melnyk P, Gaspo R, et al. Molecular mechanisms underlying ionic remodeling in a dog model of atrial fibrillation. Circ Res. 1999; 84: 776–784.
Brundel BJ, Van Gelder IC Henning RH, et al. Ion channel remodeling is related to intraoperative atrial effective refractory periods in patients with paroxysmal and persistent atrial fibrillation. Circulation. 2001; 103: 684–690.
Tessier S, Karczewski P, Krause EG, et al. Regulation of the transient outward K+ current by Ca2+/calmodulin-dependent protein kinases II in human atrial myocytes. Circ Res. 1999; 85: 810–819.