Mechanistic Inquiry into the Role of Tissue Remodeling in Fibrotic Lesions in Human Atrial Fibrillation

Abstract

Introduction

Atrial fibrillation (AF) is the most common cardiac arrhythmia associated with patient morbidity and mortality, affecting over two million people in the United States alone (1). AF is initiated when triggered activity from the pulmonary veins (PVs) propagates into atrial tissue and, via mechanisms incompletely understood, degrades into reentrant activity (2). Catheter ablation, including the electrical isolation of PVs to prevent ectopic signal propagation into atrial chambers, has emerged as a promising treatment strategy for patients who suffer from AF. However, overall success rates for terminating AF via catheter ablation are low, with recent surveys reporting only 70% success (3). Advancements in catheter ablation procedures are hindered by the elusive nature of the mechanisms underlying AF.

Recent clinical and experimental studies have convincingly demonstrated a correlation between atrial fibrosis and AF. Atrial biopsies from paroxysmal AF patients show increased interstitial fibrosis compared to sinus rhythm control subjects (4). Studies of explanted human hearts demonstrate that atrial collagen volume correlates positively with AF persistence when comparing nonarrhythmic individuals to those who suffer permanent and persistent AF (5). Furthermore, atrial fibrosis has been documented in patients who develop postoperative AF (6). In a study in which late gadolinium-enhanced magnetic resonance imaging (LGE-MRI) was used before ablation to quantify the extent of fibrosis in the left atrial (LA) wall, it was found that an increased amount of LA fibrosis is strongly associated with AF recurrence after catheter ablation (7). Although these findings convincingly show a correlation between atrial fibrosis and AF, the causal relationship between fibrosis and AF remains incompletely understood.

Fibrotic remodeling of atrial tissue involves processes that occur in parallel across multiple scales. Gap junction remodeling (GJR) at the membrane level (8), fibroblast phenotype switching at the cellular level (9), and the deposition of excess collagen at the tissue level (5) give rise to complex interactions at the organ level, setting the stage for AF initiation in the fibrotic atria. In addition, studies also indicate that the extent and distribution of fibrotic lesions within the human LA, the quantification of which has recently been made possible by improved MRI technology, may also contribute mechanistically to AF susceptibility (10). Given the complex and interdependent nature of the remodeling processes involved in fibrosis, compounded by its heterogeneous distribution throughout the human atrium, isolating the unique mechanisms by which fibrosis and its components contribute to human AF is difficult to achieve experimentally. However, the ability to obtain high-resolution images in vivo that capture both patient atrial geometry and fibrotic distribution makes it possible to reconstruct patient-specific LA geometries with accurate fibrotic lesion morphology (11). A computational model with accurate representation of atrial geometry and fibrotic lesion distribution presents a powerful tool to explore how fibrotic remodeling at the membrane, cellular, and tissue-level give rise to an arrhythmogenic substrate at the organ level.

The goal of this study was to determine the mechanisms by which fibrosis in the human atria underlies the degradation of a PV ectopic beat into AF. To achieve this goal, we used a realistic model of atrial fibrosis distribution generated from high-resolution LGE-MRI images acquired in vivo from a patient suffering from persistent AF. Modeling the fibrotic lesions incorporated representations of collagen deposition, GJR, and myofibroblast proliferation. Simulations with eight different atrial fibrosis models developed from the baseline LA geometric model were performed to determine how each component of fibrotic remodeling as well as their combinations contributes to the arrhythmogenic substrate.

Methods

Detailed Methods are available in the online Supporting Material. Briefly, a patient with persistent AF presenting for catheter ablation underwent LGE-MRI acquisition (12); a representative slice of the MR image stack is shown in Fig. 1. Image segmentation and interpolation resulted in a high-resolution image of the atrial wall with extensive fibrotic lesions (Fig. 1); the fibrotic lesion size and distribution in this patient was typical for patients with extensive fibrosis in the LA and persistent AF (12). We chose to model only the LA based on the fact that 94% of ectopic beats that result in AF initiation originate in the PVs (2) and 76% of AF sources (among which electrical rotors) were found to be located in the LA (13). A finite element tetrahedral mesh was generated from the segmented image stack, and fiber orientation was incorporated as previously described (11,14) (see the Supporting Material). Nonfibrotic regions of the tissue were represented with a modified version of the Courtemanche et al. (15) model of the human atrial action potential under AF conditions, as described by Krummen et al. (16); the model also included the formulation of the acetylcholine-activated potassium current, I_K(ACh), from Kneller et al. (17) Conductivities were chosen such that conduction velocity fell within the range recorded in the human atrium (18); additional details can be found in the Supporting Material.

Open in a separate window

Figure 1

Model generation. Representative MRI slice of the human heart, segmentation of the atrial tissue, and construction of the 3D atrial model.

Simulation protocol and data analysis

Mathematical description of current flow was based on the monodomain representation of the myocardium and simulations were executed using the simulation package CARP (CardioSolv LLC) (28). To test the arrhythmogenic propensity of the atrial models with different fibrotic representations, each substrate was paced for 5 beats at 365 ms cycle length, followed by two premature beats (S2 and S3) at cycle lengths ranging from 230 to 260 ms. Preliminary studies identified this pacing sequence as the most computationally efficient protocol to initiate AF; extending the stimulus protocol did not alter arrhythmia outcomes. Stimulus location was also chosen after preliminary results indicated that pacing from various locations in the PVs resulted in the formation of identical AF rotors, indicating that the substrate (and not pacing location) determined rotor location. Therefore, a single stimulus site was used for this study, located between the left superior and inferior PVs (Fig. 2, red dot, at top left). The initiation of reentrant arrhythmia was defined as the occurrence of unidirectional block and subsequent formation of a reentrant circuit following the premature beats. The formation of a reentrant circuit was assumed to mark the initiation of AF due to recent evidence that localized rotors or focal impulses are detected in 97% of patients with sustained AF (13).

Open in a separate window

Figure 2

Transmembrane potential maps at three time instants in three atrial models with different arrhythmogenic outcomes: neither conduction block takes place nor reentry is established (leftmost column; exclusively diffuse collagen deposition modeled in the fibrotic lesions); conduction block occurs without the formation of a reentrant circuit (middle column; GJR and diffuse collagen deposition included in the fibrotic lesions), and both conduction block and reentry occur, initiating AF (rightmost column; GJR, diffuse collagen deposition, and myofibroblasts coupled to myocytes included in fibrotic lesions). Lines of block are marked in red. Black arrows indicate direction of propagation. Fibrotic distribution is pictured at bottom right, with anatomical labels for the LA appendage (LAA), right superior PV (RSPV), right inferior PV (RIPV), left superior PV (LSPV), and left inferior PV (LIPV).

To delineate how the various components of fibrosis differentially contribute to arrhythmia propensity in the fibrotic atria, each substrate was analyzed in several ways:

• 1.
  To assess the influence of the fibrotic components on conduction, activation maps were constructed for each atrial model. Activation time was considered to be the time between S2 stimulus and local time of excitation, defined as the time at which transmembrane potential exceeded a threshold of −20 mV.
• 2.
  Myofibroblasts have a less negative resting membrane potential (V_REST) than myocytes, and thus can cause diastolic depolarization in myocytes with which they electrically couple. Indeed, research has shown that cardiomyocytes exhibit a myofibroblast density-dependent gradual depolarization when coated with myofibroblasts (29). Thus, maps of V_REST were generated for each atrial model to characterize the spatial distribution of diastolic potential and analyze its effect on conduction.
• 3.
  Experimental evidence suggests that myofibroblasts can alter the action potential duration (APD) of myocytes with which they electrically couple (30). At the tissue level, APD dispersion can increase the susceptibility of the substrate to arrhythmia (31). Therefore, APD maps for each atrial model were generated by calculating the local durations from peak action potential amplitude to 90% repolarization.

Results

Mechanism for conduction block in fibrotic lesions

The electrophysiological properties of substrates that exhibited conduction block following ectopic beats (i.e., atrial models with GJR in fibrotic regions) were compared to those of substrates that were not susceptible to conduction block (i.e., atrial models without GJR in the fibrotic regions). Slowed conduction and discontinuous propagation were observed in all models that experienced conduction block. To illustrate this finding, activation maps following a premature beat from the left PVs of select substrates are displayed in Fig. 3. Activation times following a premature beat are longer than those in sinus rhythm or following slower pacing rates due to the steep conduction velocity restitution experienced by patients who suffer from AF (32); AF patients show a conduction slowing of ∼48% with rate acceleration (33).

Open in a separate window

Figure 3

Activation maps in four atrial models, with fibrotic lesions modeled with (from left to right) diffuse collagen deposition, myofibroblast coupling, GJR, and the combination of GJR and diffuse collagen deposition.

In a substrate with only collagen distribution (diffuse or patchy) incorporated into the fibrotic lesions (left-most substrate in Fig. 3), conduction propagated in a fairly smooth and continuous manner, as indicated by the uniform and concentric activation pattern. The collagen deposition did not disrupt conduction in the model. This finding is supported by the experimental evidence that collagen of diffuse architecture (as often found in human atria (5,34)) has a minimal effect on conduction delay, even at high densities (35). Furthermore, although patchy collagen architecture at higher densities has been shown to affect conduction in the ventricles (35), existing experimental data do not indicate that the density (up to 7%) (5) of collagen of patchy architecture found in the human LA under AF contributes to conduction abnormalities.

In the substrate that contained only myofibroblasts coupled to myocytes as a representation of fibrosis (2nd from left substrate in Fig. 3), conduction through tissue proceeded discontinuously (with jagged isochrones) as it traversed more slowly through fibrotic lesions than through nonfibrotic tissue, although total LA activation time was only 5 ms longer than control. Myofibroblasts were the culprit of this slowing of conduction within fibrotic lesions, because they caused V_REST elevation in myocytes with which they were electrically coupled (29), resulting in partial sodium channel inactivation. Indeed, as indicated in Fig. 4 A, a myocyte within a fibrotic lesion experienced an elevated V_REST of −74.5 mV, compared to V_REST of −78mV within a lesion but without myofibroblasts. At the organ level, distribution of V_REST elevation in substrates with electrically coupled myofibroblasts (see Fig. 4 C, bottom row) followed the pattern of fibrotic lesion distribution (as pictured in Fig. 2). When myofibroblasts were absent from the substrate (Fig. 4 C, top row), V_REST distribution was uniform, and the corresponding activation maps had smooth isochrones.

Open in a separate window

Figure 4

(A) Action potential waveforms of myocytes within homocellular (i.e., myocyte only) and heterocellular (i.e., with electrically coupled myofibroblasts and myocytes) fibrotic lesions. (B) Difference in transmembrane potential between the two action potential waveforms of myocytes within a fibrotic lesion (homocellular lesion myocyte – heterocellular lesion myocyte), from the action potential notch to 90% repolarization. Maps of V_REST (C) and APD (D) in four models. Fibrotic lesions are modeled with (bottom row) and without (top row) myofibroblast infiltration (and coupling to myocytes), as well as with (right column) and without (left column) diffuse collagen deposition for both sets of maps. All models include GJR in the fibrotic lesions.

The substrate with only GJR in the fibrotic lesions (2nd from right substrate in Fig. 3) exhibited a similar trend of discontinuous wave propagation, resulting in jagged activation isochrones, and due again to slower conduction through fibrotic lesions. However, total LA activation time was 38 ms longer than in the control case. Thus, GJR in the fibrotic lesions caused a much larger degree of conduction slowing (total LA activation time increased by 12%) compared to that caused by myofibroblasts infiltration in the fibrotic lesions and their coupling to myocytes (total LA activation time increased by only 2%). Because PV ectopy in substrates with GJR resulted in conduction block, whereas those with only myofibroblasts did not, it follows that a large degree of conduction slowing (at least >5 ms) is necessary to support conduction block following PV ectopy in the LA. When lower values of GJR were tested and conduction slowing of 5 ms was not achieved, conduction block did not occur. When conduction was sufficiently slowed (i.e., in substrates with the GJR represented as described in the Methods section), the S3-induced wavefront encountered tissue that was both fully recovered (nonfibrotic regions), and in a refractory state (fibrotic lesions). Indeed, the red lines of conduction block in Fig. 2 correspond to the border of a densely fibrotic lesion (as seen in the fibrotic lesion distribution in Fig. 2).

Mechanism for reentry formation following conduction block in fibrotic lesions

The electrophysiological properties of the substrates in which a reentrant circuit resulted from PV ectopy, following conduction block, and degraded into sustained AF (i.e., atrial models with GJR and electrically coupled myofibroblasts in the fibrotic lesions) were compared to those of substrates that did not promote reentry despite the occurrence of conduction block (i.e., atrial models with GJR but without myofibroblasts in the fibrotic lesions) to determine the mechanisms by which myofibroblast presence and their coupling to myocytes rendered the fibrotic substrate arrhythmogenic. We investigated this by assessing the alterations in local action potential morphology (Fig. 4, A and B), as well as in organ-level electrophysiological behavior (Fig. 4, C and D).

Although myofibroblasts are inexcitable cells, they can alter action potential morphology of atrial myocytes with which they are coupled (36). Fig. 4 A shows how the transmembrane potential of a myocyte within a fibrotic lesion composed of an electrically connected network of myocytes and myofibroblasts (heterocellular lesion) compared to that of a myocyte within a fibrotic lesion with only other myocytes (homocellular lesion). Although the depolarization phase of the action potential remains relatively unaltered between the two waveforms, myofibroblasts exerted a clear influence on myocyte repolarization dynamics. The difference in transmembrane potential between the two waveforms (homocellular lesion myocyte – heterocellular lesion myocyte), from the action potential’s notch to 90% repolarization is quantified in Fig. 4 B. A myocyte within a heterocellular lesion experiences reduced peak amplitude and reduced transmembrane potential at the onset of repolarization compared to a myocyte within a homocellular lesion. The degree of the transmembrane potential reduction changes over the course of repolarization, as illustrated by Fig. 4 B. As repolarization approaches 90%, a crossover of the action potential waveforms occur (Fig. 4 A), due to the elevated resting potential of the myocyte in the heterocellular lesion. At 90% repolarization, APD is shorter in myocytes within heterocellular lesions. Indeed, APD distribution at the organ level exhibits shortening in myofibroblast-infiltrated fibrotic lesions (Fig. 4 D, bottom row, and compare to Fig. 4 D, top row). Although an average APD of 219 ms was calculated across the LA without myofibroblasts in the fibrotic lesions, the average APD was 181 ms in substrates incorporating myofibroblasts in the lesions. Introducing diffuse collagen deposition into myofibroblast-infiltrated fibrotic lesions (Fig. 4 D, bottom right compared to bottom left) exacerbated the APD shortening caused by myofibroblasts; APD decreased to 130 ms in densely fibrotic regions. This effect is because collagen, which interrupts connections between cells, limits the distance over which myofibroblasts can exert an electrotonic influence; myofibroblasts, therefore, act as a greater local current sink when collagen is present. This also explains why collagen deposition did not alter APD in substrates without myofibroblasts (Fig. 4 D, top right compared to top left). The additional APD shortening caused by the inclusion of collagen deposition (diffuse or patchy) in the fibrotic lesions was insufficient to alter the arrhythmogenic propensity of the substrates (Table 1).

Although functional coupling between myofibroblasts and myocytes is well established in the cell culture setting, direct evidence of functional gap junction formation between myofibroblasts and myocytes in the intact working myocardium is limited, having only been demonstrated in the sinoatrial node (26). However, myofibroblasts could also affect neighboring myocytes via paracrine influences. Indeed, paracrine factors released by fibroblasts have been shown to alter myocyte repolarization currents, and it has been hypothesized that these factors may contribute to the arrhythmogenic behavior of fibrotic substrates (30,37). Because the results presented herein suggest that myofibroblast-myocyte coupling alters repolarization dynamics to support reentry, we sought to determine whether direct changes in repolarization currents, as arising from paracrine effects, could also support reentrant circuit formation. Our simulation results show that atrial models in which myofibroblasts in the fibrotic lesions did not couple electrically to myocytes could also support reentry formation. This occurred when potassium currents in myocytes within the fibrotic lesions were altered. Fig. 5 presents a set of minimum paracrine alterations in the four potassium currents that was determined necessary to establish reentry formation in fibrotic models with myofibroblast proliferation but without electrical coupling to neighboring myocytes: the inward rectifier K⁺ current (I_K1), the acetylcholine-activated K⁺ current, I_K(ACh), the rapid delayed rectifier K⁺ current (I_Kr), and the slow delayed rectifier K⁺ current (I_Ks). With these paracrine effects represented in the atrial model, sustained AF also ensued following PV ectopic beats; a snapshot of atrial activity encompassing multiple wavefronts is shown in Fig. 5.

Open in a separate window

Figure 5

(A) Densities of ion currents as a function of time for a myocyte with and without an electrically coupled myofibroblast. Should paracrine effects occur, alterations in ion channel currents need to amount to these changes for reentrant circuit formation to occur without myofibroblast-to-myocyte electrical coupling. (B) Snapshot of AF activity in an atrial model where paracrine effects are represented in the fibrotic lesions (i.e., when potassium currents in fibrotic lesions were altered to mimic the changes that occur due to myofibroblast coupling).

Discussion and Conclusion

In this study, we sought to determine the mechanisms by which fibrosis and its components contribute to the degradation of PV ectopic beats into AF in the human atria. To achieve this, we used a 3D model of a patient-specific LA from a subject suffering persistent AF, capturing accurately both the atrial geometry and the distribution of fibrotic lesions. Simulations were conducted on an array of models of the same geometry/fibrosis distribution but whereby the fibrotic lesions were represented with combinations of GJR, collagen deposition, and myofibroblast proliferation, and where myofibroblasts either coupled electrically with myocytes in the fibrotic lesions or exerted equivalent paracrine influences on them. The study found that for fibrotic lesions typical of human remodeled atria under the conditions of persistent AF, the occurrence of GJR in the fibrotic lesions was a necessary but not sufficient condition for the development of AF following a PV ectopic beat. The sufficient condition was myofibroblast proliferation in these lesions, where myofibroblasts exerted either electrotonic or paracrine influences on myocytes within the lesions. Deposition of collagen in the lesions, modeled as a diffuse or patchy distribution based on experimental data, was neither a necessary nor a sufficient condition for AF development, however, it assisted the myofibroblasts’ paracrine or electrotonic effects by additionally shortening APD in the lesions.

Representing the distribution of fibrotic lesions as it occurs in the human atria under the conditions of persistent AF was essential for uncovering the specific relation between atrial fibrosis and AF in the human. Atrial LGE-MRI scans from other patients with extensive enhancement exhibit similar trends in fibrotic lesion distribution (12), lacking large contiguous regions of remodeled (fibrotic) tissue—instead, the fibrotic lesions are unevenly distributed in a noncontiguous fashion throughout the LA anterior wall, posterior wall, and the atrial septum. Although reentry development in the ventricles occurs under the conditions of myocardial infarction where a large fairly contiguous region of fibrotic scar and associated peri-infarct zone both anchor the reentrant arrhythmia and determine its anatomic pathway (38), the reentrant circuits associated with AF in the fibrotic atrium freely traverse the fibrotic regions without being anchored to any specific lesion.

This study presents a novel, to our knowledge, model of human atrial fibrosis under the conditions of persistent AF in a patient-specific representation of the atrial geometry and structure. Previous studies of fibrosis and its arrhythmogenic propensity have modeled fibrotic remodeling either by slowing conduction uniformly throughout the tissue (39) or by including only single elements of fibrosis modeled in a simplified manner, such as only fibroblasts modeled as passive resistors (40) or only collagen modeled as holes in the tissue (41). Furthermore, previous modeling efforts did not incorporate the realistic nonuniform distribution of fibrosis in the substrate (39). In contrast, our realistic model of human atrial fibrosis allows for the thorough exploration of the role of different elements of fibrotic remodeling in persistent AF.

Our finding that GJR occurrence and the resulting slowed conduction within the fibrotic lesions contributes to AF is supported by experimental evidence. When congestive heart failure, a disease strongly associated with clinical AF, was induced in dog hearts, it was found to result in discrete regions of slowed conduction and increased fibrosis compared to control (42). Furthermore, a recent study in porcine atria documented a correlation between regional Cx43 expression and sites where complex fractionated atrial electrograms (CFAE) were recorded; the expression of Cx43 at CFAE sites was significantly decreased when compared with nonCFAE sites (43).

A major finding in this study is that myofibroblast proliferation plays a key role in AF pathogenesis. Existing experimental evidence supports this finding. For one, atrial fibroblasts have a greater tendency to differentiate into activated myofibroblasts both in vitro and in vivo compared to ventricular fibroblasts (44), one explanation for why atria exhibit a more sensitive fibrotic response than ventricles in disease progression (45). Furthermore, a recent study has reported novel insights into the role of the fibroblast as the cellular link between AF and atrial fibrosis (46). In this study, it was found that transient receptor potential canonical-3 (TRPC3) channels in fibroblasts regulate cell proliferation and differentiation into myofibroblasts. The study showed that AF increases TRPC3 expression, and that TRPC3 blockade prevents AF substrate development in a dog model of electrically maintained AF, providing direct evidence of the myofibroblasts’ critical role in AF. Evidence also suggests that myofibroblasts may influence the electrophysiological substrate directly through either electrical coupling with myocytes (29,47) (which has been observed in rabbit sinoatrial node (26)), or via paracrine factors (30,37). Experimental recordings from myocytes in fibroblast-conditioned media further support our findings that altered potassium currents due to myofibroblast paracrine factors may contribute to arrhythmogenesis (37).

Considerable evidence exists suggesting a correlation between AF incidence and increased collagen (5), and it has thus been hypothesized that collagen contributes to arrhythmogenesis in the fibrotic substrate. However, in a study in which Cx43 expression was similar in young and old mice, increased collagen content in adult mice was not sufficient to promote arrhythmia inducibility (48). It has also been reported that both the density and the architecture of collagen play an important role in determining its electrical influence (35). In ventricular disease progressions or in animal models in which patchy or stringy collagen may exist at higher densities, collagen was found to likely contribute to arrhythmogenesis (35). Our results did not indicate a critical role of collagen deposition in human AF initiation following PV ectopy, but that collagen deposition did not alter the arrhythmogenic outcome resulting from other fibrotic components present in the substrate, consistent with previous simulation findings (21) and with the results of a canine study of spiral wave behavior (49). It is possible that the amount of collagen correlates with AF incidence not because of its mechanistic role in arrhythmogenesis, but because increased deposition occurs as a side effect of fibroblast proliferation and differentiation, both of which may be triggered by AF itself (46), or even by GJR (50).

The insights gained in this study regarding atrial fibrosis and its relation to AF shed important light on the reasons why approaches such as Cx gene transfer (51), APD prolongation (52), and TRPC3 channel blockade (46) prevent AF in animal studies. Furthermore, by dissecting out the mechanisms by which fibrosis leads to a generation of AF following PV ectopy the study may open new doors for the development of therapeutic targets aimed at AF termination. Given the critical importance of the electrical influence of myofibroblasts in establishing an arrhythmogenic substrate, therapies that target the intercellular communication between myofibroblasts and myocytes or that block the differentiation of fibroblasts into myofibroblasts present viable upstream targets.

Acknowledgments

References