Paradigm shifts in electrophysiological mechanisms of atrial fibrillation

Abstract

Determining the sequence of activation is a major source of information for understanding the electrophysiological mechanism(s) of atrial fibrillation (AF). However, the complex morphology of the electrograms hampers their analysis, and has stimulated generations of electrophysiologists to develop a large variety of technologies for recording, pre-processing, and analysis of fibrillation electrograms. This variability of approaches is mirrored by a large variability in the interpretation of fibrillation electrograms and, thereby, opinions regarding the basic electrophysiological mechanism(s) of AF vary widely. Multiple wavelets, different types of re-entry including rotors, double layers, multiple focal activation patterns all have been advocated, and a comprehensive and commonly accepted paradigm for the fundamental mechanisms of AF is still lacking. Here, we summarize the Maastricht perspective and Cleveland perspective regarding AF mechanism(s). We also describe some of the key observations in mapping of AF reported over the past decades, and how they changed over the years, often as results of new techniques introduced in the experimental field of AF research.

Introduction

Since the initial observations of re-entry by Meier and Mines, scientific concepts explaining the nature of cardiac arrhythmias have tremendously evolved. This is particularly true for atrial fibrillation (AF), which—even after more than a century of research on the mechanisms of the arrhythmia—still is not fully understood. Progress of technology, and discoveries of basic electrophysiological concepts are mirrored in the continuously changing view of this arrhythmia.

On the occasion of Prof. H.J.R.M. Crijns official retirement, we are delighted and honoured to be invited to summarize our views on paradigm shifts in the mechanism(s) of AF we have seen over time. We had the privilege to work with Prof. Crijns in different phases of our scientific careers, and had the opportunity to witness his enormous commitment to cardiovascular science and his exceptional understanding of cardiac arrhythmias on numerous working meetings and scientific sessions, on consensus conferences and guideline committee meetings—in which he always combined sharp scientific analysis with an elegant and friendly style of debating.

In this short review, we summarize different points of view on the pathophysiology of AF, how they evolved over the years, and how they could be integrated.

From circus movement to double layer re-entry: the Maastricht perspective

Determining the sequence of activation was always key to understand the mechanism of cardiac arrhythmias. Simple activation patterns could be determined by a limited number of electrodes or even by visual inspection of the organ in fibrillation. Also, the inducibility and maintenance of AF under various circumstances, like vagal stimulation of changes in the size of available tissue area, gave important initial hints for our understanding of AF.

Inspired by the notion of circus movement, the critical mass hypothesis, and his own experimental observations Moe formulated his Multiple Wavelet Hypothesis in the late 1950s.¹ According to this hypothesis continuous wavefront–wavetail interactions lead to wavebreak and generation of new wavefronts, while collision and fusion tend to reduce the number of fibrillation waves. As long as the number of fibrillation waves stays above a critical threshold, multiple wavelets are capable to sustain AF. Many experimental observations are in agreement with the Multiple Wavelet Hypothesis.

Experimental evaluation of Moe’s hypothesis only became possible in the early 1980s, when multi-channel amplifiers became available and computer technology emerged allowing for recording and analysis of multi-electrode recordings of complex arrhythmias. One of the pioneers in this field, Allessie et al.,² observed four to eight independent wavefronts propagating in continuously changing directions in canine atria exposed to acetylcholine obviously supporting Moe’ hypothesis. Multiple wavelets in patients with AF were also observed using ECG-imaging³ and epicardial high-density mapping of AF in patients undergoing open chest surgery confirmed highly disorganized conduction of fibrillation waves with a large variability in size and direction.^4,5 Although these observations provide circumstantial evidence for the ability of multiple wavelets to sustain AF, direct proof of Moe’s hypothesis is still lacking and alternative explanations for AF maintenance have been provided in the past decades.

Spiral wave re-entry or rotors have been identified in the atria more than 90 years ago by Garrey et al.⁶ In the 1970s Allessie et al.,⁷ interestingly before his important work in support of Moe’s Multiple Wavelet Hypothesis, demonstrated that in rabbit atria rotating re-entry does not require an anatomical obstacle. In 2000, stable microre-entrant rotors have been demonstrated to maintain AF in isolated sheep atria exposed to acetylcholine by Mandapati and Jalife.⁸

In recent years, rotors have gained a lot of popularity, particularly fuelled by the observation of Narayan and colleagues who demonstrated rotors both in the right and left atria using phase analysis of endocardial atrial electrograms.⁹ These rotors appear to be stable for up to minutes at the same location¹⁰ and ablation of the core of these rotors and other focal sources (Focal Impulse and Rotor Modulation = FIRM) led to termination of AF in >80% of patients of whom the majority was in persistent AF.⁹ Haissaguerre et al.¹¹ also demonstrated rotors, but they used ECG-imaging, a technique which allows to reconstruct conduction patterns in the heart based on an array of body surface electrograms and the individual anatomy of the thorax retrieved from CT or MRI. In this study, rotors were not spatially stable, but meandered through the atrium. Ablation of sites where they preferentially occurred resulted in termination of AF in the far majority of patients. Despite their very promising results, these studies also raise fundamental questions about the techniques used and our understanding of the role of rotors in AF maintenance in general as reviewed previously.¹²

First of all, it is currently unclear why rotors detected by FIRM are spatially stable, but those detected by ECG-imaging migrate through the atria. Secondly, both FIRM and ECG-imaging use phase analysis to determine conduction behaviour during AF. Phase analysis has been used to describe periodic phenomena not as a function of time but by its phase within a repetitive cycle. Phase analysis can be performed on signals that show a high degree of periodicity and a monophasic shape like optical action potentials or ventricular electrograms. Atrial fibrillation electrograms, however, show a high degree of fractionation and therefore need to be filtered in space and/or time to allow for a meaningful analysis of their phase.¹³ We could recently demonstrate that the combination of filtering and phase analysis leads to false identification of block lines as ‘rotors’ and therefore strongly overestimates rotational conduction behaviour of AF.¹⁴ It is therefore not a surprise that rotor ablation has not been demonstrated to be superior to PV isolation only in recent meta-analysis.¹⁵

It should also be noted that high-density, direct-contact mapping studies using traditional activation time annotation performed by several independent groups show none, a very low incidence, or a very short life span of rotational activity.^{4,5,14,16–18}

In an interesting series of papers Hansen and Federov could more recently demonstrate that AF could be maintained by stable re-entry propagating along endocardial bundle structures. Ablation at these bundles terminated AF illustrating the crucial role of these re-entries for AF maintenance.¹⁹ It should be noted, however, that these re-entrant waves propagated along anatomically preformed paths and therefore resemble anatomical re-entries rather than rotors with functional re-entry independent from anatomical obstacles. Furthermore, most of these re-entrant processes were recorded in the presence of compounds shortening the refractory period, with an AF cycle length often below 130 ms, which raises the question how representative such observations are for AF maintenance.

A very important part of the scientific legacy of the Profs. Allessie and Crijns is the notion of ‘longitudinal dissociation’ and three-dimensional conduction in the atria during AF. Using high-density, direct-contact mapping of AF they described highly disorganized fibrillation waves, often narrow and surrounded by lines of conduction block obviously oriented along the prevailing direction of atrial muscle bundles in the trabeculated network of the atrial wall.⁵ Many fibrillation waves also appeared at the surface of the epicardium as breakthroughs. These fibrillation waves were suggested to reflect transmural conduction from deeper layers of the atrial wall.¹⁷ This concept, the ‘dual-layer hypothesis’ suggests that epicardial breakthroughs originating from the endocardial bundle network provide a constant source of new wave-fronts that continue to spread on the epicardial surface and thereby contribute to the maintenance of AF. Also, this hypothesis does not assume the presence of localized drivers and for this reason is consistent with Moe’s Multiple Wavelet Hypothesis. Indeed later experimental work using simultaneous endo-epicardial mapping in the goat model of AF^20,21 and in patients¹⁸ has confirmed that the far majority of breakthroughs could indeed be traced back to fibrillation waves propagating in the endocardial bundle network. Recent computer simulation confirmed that three-dimensional conduction is expected to contribute to AF stability and becomes more enhanced as response to fibrotic alterations in the epicardial layer.²²

Many of the described discrepancies of the proposed mechanisms of AF can be explained by differences in spatial resolution, differences between endo- and epicardial recordings, patients undergoing cardiac surgery or catheter ablation, varying signal pre-processing, differences in methods for activation time annotation vs. the use of phase analysis, and various ways activation times were grouped in fibrillation waves. There remain, however, differences that are difficult to reconcile even if one takes the methodological differences into account. Crucial questions are, for example, whether repetitive activation patterns commonly occur during AF and in how they are related to sources of AF. Another unresolved issue is the enormous difference in complexity of AF with the number of fibrillation waves being estimated from just a few of them up to several hundreds simultaneously propagating through the atria. Answering these questions still appears to be pre-requisite for the identification of AF sources outside the pulmonary vein area.

The majority of the open questions concerning the basic mechanisms of AF is related to the fact that the combination of high-density and high-coverage mapping of AF has hardly been performed so far, largely for technical reasons. Besides, there is a lot to learn from shared analysis of the same mapping data using different analysis approaches. The authors invite the community to join forces and invest in such a concerted collaborative project in order to bring AF research to the next level.

From multiple wandering wavelets to localized repetitive activation sources: the Cleveland perspective

For many decades, the multiple wavelet hypothesis (multiple wandering wavelets) has been a major conceptual mechanistic model of AF. Moe and Abildskov proposed the multiple wavelet hypothesis of AF based on observations during induced AF in the canine vagal nerve stimulation model.^1,23 However, no mapping was done in these studies. According to the multiple wavelet hypothesis, AF is maintained by multiple wavelets randomly propagating throughout both atria along varying routes, primarily determined by refractory periods of the tissue. Moe et al.²⁴ developed a computational model which indicated that 23–40 such random wandering wavelets were necessary to sustain AF. Fifteen years later, data from studies of induced AF in a Langendorff-perfused, acetylcholine infused, in vitro canine atrial model by Allessie et al.²⁵ supported the Moe hypothesis, but indicated that only 4–6 simultaneously circulating random, wavelets were necessary to sustain AF. However, our recent mapping studies of the original canine Moe/Abildskov model showed that in contrast to their prediction, independent repetitive focal sources initiated wavefronts reactivating the atria, thereby explaining the mechanism maintaining the Moe/Abildskov model of AF.^26,27 All wavefronts emanated from focal sources, largely either colliding or merging with another wavefront at variable sites, or encountering refractory tissue. In short, there are obvious differences between these two studies. Four of the major differences are (i) in vitro vs. in vivo studies, (ii) endocardial, high-density mapping vs. epicardial, higher density mapping, (iii) sequential uni-atrial mapping vs. simultaneous bi-atrial mapping, and (iv) high-dose acetylcholine infusion in the Langendorff-perfused atria vs. vagal nerve stimulation per Moe and Abildskov. Perhaps the latter difference is most important, because the distribution of acetylcholine is diffused in the Langendorff-perfused model and localized at the fat pads in the vagal nerve stimulation model. Also, the concentrations of acetylcholine were almost certainly very different than the effects of vagal nerve stimulation. Thus, the effects of acetylcholine on the atria likely were physiologically quite different between the two models. Therefore, the Allessie et al. interpreted their data as confirming the multiple wandering wavelets is quite a different model than that studied by Moe/Abildskov. Although Allessie et al. studied a different canine model than our studies, both studies reported (i) a similar number of wavefronts, (ii) similar activation patterns, and (iii) periods of electrical atrial silence reactivated by focal activation. Allessie et al. postulated that an electrically silent atrium was reactivated by a wavefront propagating from the other atrium due to their inability to map both atria simultaneously. However, in our bi-atrial, simultaneous mapping studies (510 electrodes), each electrically silent atrium was reactivated by a distinct focal source with QS morphology, and was not reactivated by wavefronts propagating from the other atrium.

Over the past 30 years, with improved mapping technologies applied to electrophysiological studies of AF in animal models and patients, a repetitive activation pattern (driver mechanism) of AF has been demonstrated. In in vitro models of AF using cholinergic stimulation, Schuessler et al.²⁸ demonstrated in a right atrial preparation that, with increasing acetylcholine concentrations, activation patterns characterized by multiple wandering wavelets converted to a repetitive re-entrant circuit with rapid and regular activation that resulted in fibrillatory conduction. Also, Jalife and Mandapati et al.,⁸ using optical mapping in a sheep atrial model perfused with acetylcholine, demonstrated a rapid, repetitive, rotating activation (stable rotor) acting as a high frequency source (driver), that caused fibrillatory conduction. In addition, a recent study by Peters and colleagues using an optically mapped canine cholinergic AF preparations demonstrated that the predominant conduction pattern during AF was repetitive focal activation.²⁹ A study by Fedorov’s group in in vitro explanted human right atrial AF preparations has recently shown that repetitive transmural re-entrant activation around atrial fibrosis drove AF, and was subsequentyly ablated to terminate AF.¹⁹ Interestingly, an epicardial repetitive focal activation pattern was always observed at the location of transmural re-entry around the atrial fibrosis. In an in vivo canine model of AF, our high-density mapping study using either the canine sterile pericarditis or congestive heart failure model demonstrated that a repetitive re-entrant circuit or focal source was a predominant mechanism sustaining AF.^30,31

In patients with persistent and long-standing persistent AF, there is enough evidence that AF is driven by localized repetitive activation sources in the atria, such as focal sources or different types of re-entry, including rotors.^32–42 Recently, ablation of localized repetitive activation sources in patients with persistent and long-standing persistent AF resulted in AF termination in some patients,^35–42 although some endocardial mapping data have recently been challenged.^14,43–52 In several epicardial mapping studies in patients with persistent and long-standing persistent AF, a focal source was identified in the left atrium and/or right atrium,^53–56 and repetitive activation patterns were found in the left atrium,^57–60 suggestive of a driver mechanism. Our recent studies using simultaneous, bi-atrial, high density (510–512 electrodes), epicardial mapping in patients with persistent and long-standing persistent AF found that wavefronts emanating from focal sources and breakthrough sites maintained AF.^33,34 These focal sources manifested a QS uni-polar morphology suggesting that the ectopic focal discharges were the underlying mechanisms. The focal sources were either sustained over the duration of the period of mapping analysis (32 s), or intermittent. Periods of focal activity were associated with pre-mature activation of the focal site by another earlier wavefront, or the focal site simply paused temporarily, and was passively activated. Also, repetitive focal activations sometimes generated wannabe re-entry, i.e. incomplete re-entry, in areas of slow conduction, but no actual re-entry was found. Other high-density sequential area epicardial mapping studies in patients with persistent and long-standing persistent AF demonstrated only intermittent focal activation,^17,18 or intermittent focal activation and re-entry.⁴ Using endocardial catheter mapping in paroxysmal and/or persistent AF, sustained and intermittent focal sources were found by Haïssaguerre’s group.³² Using phase analysis, Narayan and colleagues³⁵ reported finding putative focal drivers and rotors with spatial stability. However, recent studies from Narayan’s group⁴⁰ and Schotten’s group¹⁴ demonstrated that phase analysis produces a different mechanistic result than classical activation sequence analysis. Also, recent studies by Chauhan and colleagues³⁷ in patients with persistent AF identified focal sources with QS morphology, and ablation of them terminated AF. A non-invasive mapping study by Rudy’s group³ showed that in paroxysmal and persistent AF patients, the most common activation pattern consisted of multiple 2–5 concurrent wavelets with simultaneous focal activation from areas near the pulmonary veins (69%) and non-pulmonary veins (62%). Re-entry was seen rarely, and was rarely sustained > 1 rotation. Also, in a non-invasive mapping study in patients with persistent and long-standing persistent AF, Haissaguerre’s group¹¹ identified driver mechanisms (80.5% re-entry and 19.5% focal) that recurred repetitively at the same region.

In summary, using different types of mapping technologies and analysis algorithms, targeted ablation of localized repetitive activation sources during AF have been successfully performed to achieve acute termination of AF.^35–42 Although there is a controversy regarding reported predominant activation patterns (rotational vs. focal), neither repetitive activation patterns supports the multiple wavelets hypothesis.

Conclusion

Clearly, the mechanism(s) of AF is/are not fully understood. Theories for the maintenance of AF in principle are not mutually exclusive. It is possible that several mechanisms of AF might exist, even in the same patient. Hypotheses suggesting localized drivers are relatively easy to validate by eliminating the localized source. However, the targeted ablation approach using identification of repetitive activation patterns distinctive for an AF source is extremely dependent on high inter-electrode spatial resolution for accuracy. Also, understanding activation sequences precisely in real-time is challenging. Hypotheses suggesting chaotic forms of AF without localized drivers, e.g. the multiple wavelet hypothesis, are notoriously difficult to prove in patents. In fact, as long as we cannot detect all electrical activity in the entire atria, it will remain difficult to provide evidence for Moe’s Multiple Wavelet Hypothesis but that this does not exclude the possibility that multiple wavelet can sustain AF. Finally, we look forward to studies of that fully describe the mechanism(s) of AF, their treatment, and better still, their prevention.

Funding

This work was supported in part by grants from R01 HL146463 from the National Institutes of Health, National Heart, Lung, and Blood Institute; and Elisabeth Severance Prentiss Foundation to A.W. This work was further supported by Health∼Holland, Top Sector Life Sciences & Health (EXACT-IN-AF, PPP allowance LSHM19017 to S.Z.). Health∼Holland, Top Sector Life Sciences & Health (EXACT-IN-AF, PPP allowance LSHM19017 to S.Z., by the ITN PersonalizeAF, Horizon 2020 research and innovation programme PersonalizeAF under the Marie Skłodowska-Curie grant agreement No.860974 to US, and by the Netherlands Heart Foundation (CVON2014-09, RACE V: Reappraisal of Atrial Fibrillation: Interaction between hyperCoagulability, Electrical remodelling, and Vascular Destabilisation in the Progression of AF) to US.

Conflict of interest: U.S. received honoraria or consultancy fees from EP Solutions, Roche, Johnson & Johnson and research grant from Roche and EP Solutions. U.S. is a shareholder of YourRhythmics BV. All other authors have no conflicts of interest.