Cardiac dynamics: Alternans and arrhythmogenesis.

Pre-existing heterogeneities present in cardiac tissue are essential for maintaining the normal electrical and mechanical functions of the heart. Exacerbation of such heterogeneities or the emergence of dynamic factors can produce repolarization alternans, which are beat-to-beat alternations in the action potential time course. Traditionally, this was explained by restitution, but additional factors, such as cardiac memory, calcium handling dynamics, refractory period restitution, and mechano-electric feedback, are increasingly recognized as the underlying causes. The aim of this article is to review the mechanisms that generate cardiac repolarization alternans and convert spatially concordant alternans to the more arrhythmogenic spatially discordant alternans. This is followed by a discussion on how alternans generate arrhythmias in a number of clinical scenarios, and concluded by an outline of future therapeutic targets for anti-arrhythmic therapy.

Introduction

Pre-existing heterogeneities present in cardiac tissue are essential for maintaining the normal electrical and mechanical functions of the heart. However, an increased risk of cardiac arrhythmias can result from the exacerbation of such heterogeneities, which can occur under pathological conditions or following the administration of cardiotoxic drugs. The emergence of dynamic factors, which can interact with each other as well as with pre-existing tissue heterogeneities, can produce arrhythmogenic repolarization alternans and therefore cardiac arrhythmias. The focus of this review is to illustrate the mechanisms that (i) generate cardiac alternans, (ii) convert spatially concordant alternans to the more arrhythmogenic spatially discordant alternans, and (iii) are responsible for the production of arrhythmias in a number of clinically relevant conditions. This is concluded by a discussion on future therapeutic targets for anti-arrhythmic therapy.

Cardiac alternans

Cardiac alternans are beat-to-beat oscillations in either arterial pulse or electrocardiographic QRS and T waves. Of these, T-wave alternans (TWAs) have been associated with re-entrant arrhythmogenesis and identified as a good predictor of sudden cardiac death [1]. They are due to alternations in repolarisation time courses (measured as action potential durations, APDs) at the cellular level, which increase in amplitude with faster heart rates. TWAs have been observed in a number of conditions, including electrolyte abnormalities, hypothermia, coronary artery disease, post-myocardial infarction, long QT and Brugada syndromes, vasospastic angina, dilated, hypertrophic, and Takotsubo cardiomyopathies, and end-stage heart failure.

APD restitution-dependent mechanisms

The relationships between the diastolic interval (DI), APD, and basic cycle length (BCL) are shown in Fig. 1. BCL is the sum of APD and DI. The mechanism of APD alternans was first described by using a graphical method, relating them to APD restitution [2]. This refers to the normal shortening of APD in response to faster heart rates, and is thought to be an adaptive mechanism for preserving diastole at such rates. It can be defined as the dependence of APD on the previous DI. Experimentally, this can be determined by using an S1S2 protocol, which gradually shortens the interval between the S1 and S2 stimuli, or by using a dynamic pacing protocol, which increases the heart rate by progressively reducing the BCL. While both methods can be used to measure APD restitution [3], the S1S2 restitution curve is a measure of the immediate response to a change in BCL, whereas the dynamic restitution curve is a measure of the steady-state response [4]. Fig. 2 shows a typical APD restitution curve obtained from mouse hearts, APD=f (DI), where f is the function relating the new APD to its previous DI. The dashed line indicates the gradient of the curve and the gray area refers to values of DIs with gradients greater than one.

Fig. 1

Voltage trace showing the relationships between action potential duration (APD), diastolic interval (DI), and basic cycle length (BCL).

Fig. 2

An APD restitution curve describes the relationship between the APD and the previous diastolic interval (solid line). The gradients of the curve are represented by the broken line. The values of DIs at which such gradients are greater than one are represented by the gray box.

The gradient of the restitution curve is a collective measure of the recovery of all the ion channels opened during the cardiac action potential. First, of these channels, sodium channels recover from inactivation rapidly, and therefore their effects on APD restitution occur mainly at short DIs, between 0 and 40 ms in human hearts. However, if the recovery of sodium channels is slowed, which can occur under ischemic conditions [5], their effects on APD restitution would be extended to longer DIs. Second, the L-type calcium channels recover more slowly than sodium channels, and their effects are therefore observed in the short and intermediate DI ranges, between 0 and 100 ms. These calcium channels provide the majority of the inward current during the plateau phase of the action potential and therefore exert major effects on APD restitution. Their inhibition leads to reduced gradients of APD restitution curves. Third, time-dependent potassium channels, such as the voltage-gated delayed rectifiers, show the slowest recovery compared to other ion channels and therefore their effects are observed over a much larger DI range beyond 100 ms. In addition, the block of potassium channels shows reverse use dependence, where there is less block with increasing use [6]. Thus, the block increases during phase 4 of the action potential (diastole) and decreases during the plateau phase. Consequently, potassium channel blockers, which prolong APDs, have greater effects at long BCLs (bradycardia) and long DIs (e.g. compensatory pause after an ectopic beat), but have much smaller effects at short BCLs (tachycardia) and DIs [6]. They generally increase the gradients of APD restitution curves. The steep portion of the APD restitution curve is relevant in sinus tachycardia, where the heart rate is increased. It is also relevant in heart failure or the congenital and acquired long QT syndromes. In these conditions, APD is prolonged and therefore the DIs can become short enough to engage the steep portion of the APD restitution curve even at normal heart rates.

Fig. 3 shows cobweb plots that can be used to determine the stability of APD alternans. As BCL decreases, APD also decreases and the relationship BCL=APD+DI can be shown graphically as a straight line with a gradient of −1. The equilibrium point of APD for each BCL is the intersection point of the restitution curve and this line, which has the coordinates [DIs, APDs]. The stability of APD can be determined by perturbing the DI by a small amount, δ, such that DI=DI+δ. A negative δ that shortens the DI would move DI to the left. This will in turn produce a shorter APD. However, this will result in a long DI and therefore a long APD.

Fig. 3

APD restitution curve plotting APD against the previous DI (solid line) along with their gradients (broken line). The values of DIs with gradients greater than one are represented by the gray box. The cobweb plot shows that when the APD restitution gradient is less than one, a stable equilibrium point is produced on successive beats.

If the gradient of the APD restitution curve at the intersection with the BCL line is less than one (Fig. 3), then the alternans are transient and will return to the stable equilibrium point over subsequent beats. However, if this gradient is greater than one (Fig. 4), then the amplitude of the alternans will increase, eventually leading to 2:1 block in this simplified situation where the APD restitution curve is linear. As indicated in the figure, the slope of the APD restitution curve decreases with longer DIs and therefore alternans will eventually reach a maximum stable value because the flat restitution curve prevents their growth.

Fig. 4

APD restitution curve plotting APD against the previous DI (solid line) along with their gradients (broken line). The values of DIs with gradients greater than one are represented by the gray box. The cobweb plot shows that when the APD restitution gradient is greater than one, an unstable equilibrium point is produced on successive beats, eventually leading to conduction block.

APD restitution-independent mechanisms

Restitution analysis assumes that alterations in DIs produce the changes in APDs, and that the gradient of the restitution curve determines the extent to which alterations in DIs produce APD alternans. However, a feedback-based protocol that permits explicit control of DI independent of APD has subsequently been developed [7], [8]. Experiments using such a protocol have demonstrated APD alternans can occur during constant DI pacing [9]. This means that DIs do not necessarily have to change to produce APD alternans, which therefore have APD restitution-dependent and restitution-independent components. Therefore, it can be concluded that other factors influence this situation. Some of these factors that have been identified include cardiac memory, calcium handling dynamics, ventricular effective refractory period (VERP) restitution, and mechano-electric feedback, which will be discussed in turn (Table 1).

Table 1

APD restitution-dependent and APD restitution-independent mechanisms for producing APD alternans.

APD Alternans	Explanation	References
APD restitution-dependent	APD is dependent on the previous DI. Abrupt change in DI leads to engagement in the steep portion of the APD restitution curve.	[2]
APD restitution-independent:		[9], [71]
Cardiac memory	APD depends on not only the preceding DI but a series of DIs preceding it, i.e. the pacing history is important, which is termed APD accommodation.	[3], [11]
	Rate-dependent memory, termed hysteresis, results in persistence of alternans despite subsequent slowing of heart rate.
Calcium handling	Ca2+→APD coupling	[14], [15], [16], [17], [18]
VERP restitution	VERP can diverge from APD, e.g. during hypokalemia.	[19], [63]
Mechanoelectric feedback	Mechano-sensitive ion channels can influence the membrane potential	[21], [22]

Cardiac memory results from the slow recovery of ion channels as well as the gradual accumulation of ions on either side of the cell membrane. Consequently, the pacing history is also important, and so APD depends not only on its immediately preceding DI but also on the series of DIs preceding it, i.e., APD=f (DI, DI, DI…). Second, APD depends not only on the preceding DIs but also on the previous APDs. Therefore, in the presence of these short-term memory effects, known as APD accommodation [10], APD restitution gradients greater than one may not predict the onset of APD alternans [3], [11]. Nevertheless, better methods have been developed to investigate the rate- and memory-dependent aspects of APD restitution. For example, the restitution portrait was generated by modifying the pacing protocol to include a measurement of the transient response of the APD after a change in BCL [4]. This allows measurements of various aspects of APD restitution simultaneously, including dynamic restitution, S1S2 restitution, and the transient response that arises from short-term memory. Such a protocol has subsequently been used to examine restitution properties in the ventricles of perfused whole hearts [12]. Moreover, rate-dependent memory, known as hysteresis, has been demonstrated for repolarisation alternans observed in humans [13]. This refers to alternans induced by rapid pacing that persist despite subsequent slowing of the heart rate [14]. Theoretical work has shown that both restitution-dependent and independent factors are responsible, and abnormal calcium handling plays an important role [14]. Studies investigating onset of alternans should specify the initial pacing conditions to account for hysteresis.

Calcium handling dynamics, when altered, can also produce APD alternans [15]. Membrane potential and [Ca2+] are bi-directionally coupled. ICa,L is a major determinant of both APD and [Ca2+]. If APD alternates, [Ca2+] will also alternate in response to the alternating ICa,L amplitude (APD→Ca coupling). A longer DI allows a longer recovery time of the L-type calcium channels, which would increase the SR calcium load and therefore the calcium transient amplitude. Conversely, [Ca2+] also affects APD through its effects on the calcium-sensitive currents, such as ICa,L, INCX, INS, ICl,Ca, that are active during the plateau phase (Ca→APD coupling). The situation is complex here because [Ca2+] can have opposing effects on APD. It shortens APD by potentiating the calcium-induced inactivation of ICa,L but prolongs APD by increasing the inward current produced by INCX. The net effect of [Ca2+] on APD depends upon which of the two factors dominates. A larger Ca2+ transient causing prolonged and shortened APDs is termed positive and negative coupling, respectively [16]. Furthermore, the propensity for pacing-induced calcium alternans increases with increasing mitochondrial dysfunction through either dissipation of the mitochondrial membrane potential or inhibition of ATP synthesis [17]. What is clear is that APD alternans is an emergent property involving both voltage- and calcium-dependent mechanisms. Recent experiments using stochastic pacing demonstrated the use of a novel parameter, λalt, to reveal the onset of alternans as well as to distinguish between voltage- and calcium-driven APD alternans [18].

In situations where the VERP diverges from APD, as has been shown in hypokalemia, restitution gradients may not accurately predict the onset of alternans [19]. In this case, the gradients of VERP restitution curves may be a better indicator. An additional advantage of VERP over APD restitution is that VERP can be measured without the need to accurately record action potential waveforms from monophasic action potential (MAP) recordings.

Moreover, mechanical contraction can modulate the electrical activity of myocytes, a phenomenon referred to as mechano-electric feedback (MEF). Electrical alternans have been observed during simulated pulsus alternans produced by clamping the aorta on alternate beats [20]. In this situation, the myocytes of the vessel wall did not actually contract and therefore changes in [Ca2+] probably did not cause these electrical alternans. It instead pointed to mechano-sensitive ion channels [21], such as volume- or stretch-activated channels (SACs), being responsible. Indeed, these mechano-sensitive ion channels can influence the membrane potential on a beat-to-beat basis and may therefore influence cardiac dynamics. Thus, preliminary evidence shows that SAC activation can have suppress spatially concordant alternans but exacerbate discordant alternans [22].

Spatially concordant and discordant alternans

APD alternans can be either spatially concordant or discordant (Table 2). As the pacing rate is increased, spatially concordant alternans are observed, in which APD is long throughout the cardiac tissue on one beat and short on the next beat, i.e., APDs in different regions alternate in phase with each other. When the pacing rate is increased further, spatially discordant alternans are produced, in which APD is long in one region but short in an adjacent region, and changes phase on the next beat; i.e., APD in different regions alternate out of phase with each other. A number of mechanisms have been identified as being responsible for the production of spatially discordant APD alternans. These can involve pre-existing heterogeneities, which often interact with dynamic factors to produce them. However, pre-existing tissue heterogeneities may not be necessary; the presence of dynamic factors alone may be sufficient for producing spatially discordant alternans [23].

Table 2

Mechanisms for producing spatially concordant and spatially discordant APD alternans.

Alternans	Mechanisms	Clinical relevance	References
Spatially concordant alternans (SCAs)	Positive Ca2+→APD coupling		[40]
Spatially discordant alternans (SDAs):
Pre-existing tissue heterogeneities	Steep APD restitution	AF, Long QT syndromes, Brugada syndrome, heart failure, hypothermia	[24], [62], [72]
	Spatial gradients in Ca2+transients	Heart failure	[27]
	Spatial gradients in VERP	Hypokalemia	[40], [73]
	Gap junction uncoupling or downregulation	Oculodentodigital dysplasia, Naxos disease	[28], [29], [74], [75]
	Abnormal sodium channel function	Heart failure, Long QT syndrome type 3	[31]
	Fibrosis	Heart failure	[30]
Dynamic factors	CV restitution	Ischemia, sodium channel blockade, hypothermia	[5], [34], [62]
	Steep VERP restitution	Hypokalemia	[38]
	Steep relationship between sarcoplasmic reticulum calcium release and diastolic sarcoplasmic reticulum calcium load	Heart failure, exercise, catecholaminergic polymorphic ventricular tachycardia, atrial fibrillation	[76]
	Negative Ca2+→APD coupling	Heart failure	[40], [38]
	Calcium accumulation in the sarcoplasmic reticulum	Heart failure, exercise, catecholaminergic polymorphic ventricular tachycardia	[41]
	Reduced repolarization reserve	Long QT syndromes	[47]
	After-depolarization phenomena	Atrial fibrillation, heart failure	[77], [78]
	Ectopic beats	Heart failure, long QT syndromes	[34]
	Sympathetic stimulation	Heart failure, exercise	[61]

Pre-existing tissue heterogeneities

Pre-existing tissue heterogeneities are normally present in the heart, such as spatial gradients in repolarization between the endocardium and epicardium or the base and apex. A steep APD restitution gradient can convert spatially concordant alternans to discordant alternans [24]. Like APD, Ca2+ handling in the ventricle also exhibits apex-base [25] and endocardium-epicardium [26] gradients. Since Ca2+ affects APDs, heterogeneity in Ca2+i handling can produce spatially discordant APD alternans via either differential Ca2+→APD coupling or spatial heterogeneities in the phase of the Ca2+ alternans [27]. Moreover, electrotonic coupling between cardiomyocytes via gap junctions attenuates differences in properties between individual cells. When gap junctions are uncoupled, pre-existing heterogeneities such as spatial gradients in APDs, CVs, and Ca2+ can become amplified, leading to discordance [28], [29]. Finally, under conditions where cardiac conduction is abnormal, e.g. in heart failure [30], the threshold for inducing spatially discordant alternans is lower. This may be due to fibrosis, altered expression of gap junctions, or delayed recovery of sodium channels from inactivation [31].

Dynamic factors

A number of dynamic factors can convert spatially concordant to discordant APD alternans. First, CV restitution describes the relationship between the CV and the preceding DI (Fig. 5). It is almost entirely dependent on the recovery of sodium channels because the upstroke of the action potential is mainly determined by these channels [32]. Under normal conditions, these channels recover from inactivation rapidly, and therefore CVs only slow at very short DIs, between 0 and 10 ms (Fig. 5, solid line). However, if their recovery is slowed, CVs vary over a broader range of DIs (Fig. 5, dashed line). Fast pacing can engage the steep portion of the CV restitution curve to convert pre-existing spatially concordant alternans to discordant APD alternans when the tissue size is sufficiently large [33]. Initially, at a relatively low heart rate, spatially concordant APD alternans are observed. When the heart rate increases, a beat with a long APD has a sufficiently short DI after it. This engages CV restitution, causing the CV of the subsequent beat to decrease. The conduction slowing allows the DI to increase slightly, which in turn allows APD to increase slightly. This process amplifies itself with subsequent beats, eventually producing discordant alternans [34]. This mechanism may be important during ischemia or sodium channel blockade where sodium channel recovery is slowed [5], [35]. This would extend the range of DIs over which CV varies, producing spatially discordant APD alternans even at normal heart rates [36], thereby increasing arrhythmogenic risk [32], [37] in both of these conditions. Alternatively, CV restitution can also interact with steep VERP restitution or altered [Ca2+] cycling dynamics to produce discordant alternans [38](Fig. 6).

Fig. 5

CV restitution curve. CV is plotted against its previous DI in which sodium channel recovery is normal (solid curve) or slowed (dashed curve). The latter is observed in clinical situations such as tissue ischemia, after application of sodium channel inhibitors and hypothermia.

Fig. 6

Pre-existing heterogeneities or dynamic factors can convert spatially concordant APD alternans to spatially discordant APD alternans.

Ca2+alternans can be induced in cardiac cells by making sarcoplasmic release of calcium strongly dependent upon the SR Ca2+content [39]. They can become discordant if Ca2+→APD coupling is changed from positive to negative [40]. Recently, a novel mechanism for the production of spatially discordant alternans was demonstrated despite positive Ca2+→APD coupling [41]. Here, rapid pacing led to Ca2+accumulation in the sarcoplasmic reticulum over many beats, leading to Ca2+alternans that were out-of-phase with each other [41]. Compelling evidence suggests that APD alternans, whether in the atria or the ventricles, are driven by Ca2+ alternans [42], which are attributable to ryanodine receptor refractoriness [43].

The concept of repolarisation reserve was introduced to explain the occurrence of arrhythmias, positing that redundancy in repolarization exists [44], [45]. The idea is that elimination of one repolarizing current need not necessarily lead to failure of repolarisation because of compensation by other repolarizing currents [46]. Recent experiments in transgenic LQT1 rabbits have shed light on the mechanistic link between repolarisation reserve and alternans [47]. Interestingly, in this model, reduced repolarization reserve per se was paradoxically associated with a shallower gradient of APD restitution and a higher threshold for inducing alternans. It was only tachycardia pacing that led to steepening of the restitution curves and the development of alternans. This in turn has been attributed to abnormal calcium handling, in which IKs downregulation led to discordant Ca2+alternans. Under conditions of prolonged repolarization or abnormal calcium handling, after-depolarization phenomena are observed and these can produce spatially discordant alternans. Where the voltage change brought about by such phenomena are sufficiently large, an ectopic beat can be generated, which can cause dispersion of the diastolic interval, thereby converting spatially concordant alternans to discordant alternans [34].

Alternans and arrhythmogenesis in different clinical conditions

APD alternans, whether spatially concordant or discordant, can produce arrhythmias. For example, spatially concordant alternans themselves can produce 2:1 conduction block, thereby initiating re-entry [48]. Nevertheless, discordant alternans are considered to be more arrhythmogenic. They can produce large spatial gradients in repolarisation and refractoriness, which can result in local conduction block of a premature extrasystole (such as a premature ventricular complex, PVC [37]), thereby facilitating re-entry [23], [49]. They can also promote phase 2 re-entry involving either a fixed or a variable Ito, allowing antegrade and retrograde phase 2 re-entry, respectively [50].

These mechanisms for producing spatially discordant alternans are important in a number of clinical conditions, potentially causing lethal arrhythmias. In heart failure, there is extensive ion current remodeling with a lower threshold for inducing APD alternans [51]. APDs are prolonged due to downregulation of potassium currents and increased late sodium current [30]. This would allow the steep portion of APD restitution to be engaged, and therefore generation of spatially concordant alternans. These can become discordant in the presence of conduction abnormalities [30], cardiac fibrosis [52], or abnormal Ca2+ handling dynamics, such as increased SR Ca2+ leak [53], decreased SERCA pump activity [54], increased NCX currents [55], or steeper fractional release of SR Ca2+ content [56]. In catecholaminergic polymorphic ventricular tachycardia (CPVT), mutations in the ryanodine receptor lead to diastolic calcium leak and generation of discordant Ca2+ alternans [57], [58]. Long QT syndromes are characterized by APD prolongation, a reduction in repolarization reserve, and increased APD restitution gradients leading to the production of APD alternans. Spatial heterogeneities in repolarization are exacerbated due to differences in ion channel expression and heterogeneities in restitution across the myocardial wall.

Other pro-arrhythmic conditions are clearly associated with a flatter APD restitution curve, but discordant alternans can be generated when regional differences in restitution lead to spatial heterogeneities in APDs [59]. In myocardial ischemia and sodium channel blockade, CV restitution may be more important in the conversion of concordant alternans to discordant alternans [60]. Increased beta adrenergic drive, which can occur in heart failure or exercise, can increase the maximum gradients of APD restitution and produce discordant alternans [61].

Regardless of the mechanisms generating these arrhythmogenic alternans, the final common pathway involves wavebreak, conduction block, and the initiation and maintenance of re-entrant arrhythmias [62]. It should be recognized that alternans are only one factor in determining arrhythmogenesis. An anti-arrhythmic state can occur even in the presence of both steep APD restitution and discordant alternans, as exemplified by hypokalemia [19]. Heptanol, a gap junction uncoupler, was shown to exert anti-arrhythmic effects in hypokalemia by influencing VERP alone [63]. This finding is perhaps surprising, given that reduced electrotonic coupling should exacerbate dispersion in APDs and promote arrhythmogenesis.

Future therapies

The question of how understanding the cardiac dynamics can enable us to devise better pharmacotherapy for arrhythmia management persists. Numerous studies using animal models have demonstrated that the anti-arrhythmic actions of many drugs are in part mediated by their effects on cardiac dynamics. These include traditional agents such as beta-blockers as well as novel drugs such as late sodium current blockers and gap junction openers [64], [65], [66] (Fig. 7). Gap junction inhibitors can exert anti-arrhythmic effects by prolonging effective refractory periods [63], [67]. Moreover, mild loss of gap junction function in non-uniform tissue may paradoxically increase CV and improve the safety margin of conduction [68]. This in turn could remove unidirectional conduction blocks, converting them into bilateral conduction [69]. In contrast, gap junction enhancers can improve conduction and reduce the spatial heterogeneities in repolarization and refractoriness, thereby suppressing discordant alternans and arrhythmogenesis. Their effects on calcium dynamics are complex, depending on the nature of Ca2+→APD coupling. The differences in Ca2+ transient between adjacent cells are amplified in the case of negative coupling, but reduced with positive coupling [40]. Activation of stretch-activated channels exerts opposing effects on alternans, suppressing those that are concordant whilst exacerbating those that are discordant [22]. SAC inhibitors could potentially exert anti-arrhythmic effects by suppressing discordant alternans. Discordant alternans can also be inhibited by late sodium channel blockers [65], ryanodine receptor stabilizers [57] or anti-fibrotic agents [70], which would prevent conduction block and inhibit arrhythmogenesis. Some of the examples described above illustrate the difficulty in predicting the overall electrophysiological effects of a drug, and some may exert their anti-arrhythmic effects without influencing restitution [71], [72]. Future efforts therefore require a computational strategy in which modeling of the heart can be achieved at different levels of biological organization to take into account of the complex spatiotemporal properties of cardiac dynamics.

Fig. 7

Potential molecular targets for anti-arrhythmic therapy, by suppression of spatially discordant APD alternans by influencing cellular and tissue dynamics.

Conflict of interest

All authors declare no conflict of interest related to this study.