Can a computational model explain the role of sex hormones in atrial fibrillation?

1. Introduction

Atrial fibrillation (AF) is characterized by rapid and uncontrolled atrial activity. The initiation and maintenance of AF stem from the interaction between a trigger, such as a rapidly firing focus, and a substrate, such as the mechanical and anatomical characteristics of the atria. Fibrosis, defined as the excess and persistent deposition of extracellular matrix (ECM) proteins, is a key substrate that promotes AF by creating nonconductive barriers and disrupting cell-to-cell electrical coupling, producing heterogeneous conduction that favors reentry. AF, in turn, promotes further fibrosis through neurohormonal activation, inflammation, and mechanical stretch, reinforcing a self-perpetuating cycle in which AF begets AF. Because fibrosis both drives and is driven by AF, it represents a critical node for interrupting this cycle and a candidate marker for guiding AF ablation therapy.

Biological sex and age are important modifiers of fibrotic remodeling and AF. Men are associated with a higher risk of developing AF, while women experience higher rates of hospitalization and arrhythmia recurrence after ablation, alongside a greater degree of fibrotic remodeling. AF prevalence rises steeply with age, coincident with declining sex hormone (estrogen, E2; testosterone, TS) levels in both sexes, and aging-associated ECM remodeling parallels these hormonal transitions during menopause and andropause. This temporal overlap suggests that sex hormone-dependent regulation of fibrotic signaling may critically shape AF susceptibility across the lifespan. Extensive experimental evidence supports this idea: both E2 and TS confer protection against fibrotic remodeling, limiting ECM production and fibroblast activation in response to AF-relevant stimuli. However, the mechanistic basis of sex hormone-dependent antifibrotic protection have remained poorly understood. In this issue, Khorasani and colleagues use computational modeling to provide mechanistic insights into how E2 and TS impact fibrotic signaling in cardiac fibroblasts¹.

2. Computational Models Can Provide Mechanistic Insights

Cardiac injury triggers fibroblasts to be differentiated into myofibroblasts, which overexpress and deposit excessive ECM components. This differentiation is driven primarily by transforming growth factor beta (TGFβ) and angiotensin II (AngII). Sex hormones, such as E2 and TS, have been shown to inhibit fibroblasts differentiation. E2 inhibits AngII-induced fibroblast differentiation^2,3, while TS attenuates TGFβ and AngII-enhanced fibroblasts activities⁴. While the impact of sex hormones on isolated pathways has been studied, the cross talk between pathways is challenging to study experimentally since fibroblast activation consists of a complex signaling network influenced by inflammatory, neurohormonal, and mechanical regulators.

Computational models for fibroblasts have been developed in the past to address this issue, using logic-based Hill models that integrate prior knowledge of individual pathways to study system-wide behavior. Zeigler et al. were the first to apply this framework to construct a cardiac fibroblast signaling network aimed at understanding the mechanisms behind myofibroblasts differentiation⁵. Watts et al. became the first to integrate sex hormones in a fibroblast computational model by expanding Zeigler’s model to include E2, and established a method for encoding sex-specific physiology by adjusting receptor activity and hormone levels across male, premenopausal and postmenopausal models⁶. Watts’ extended model provided insights on why premenopausal women generally develop less fibrosis than men and how E2 influences fibrotic pathways, but it did not capture the effects of TS. In parallel, Khorasani et al. extended Zeigler’s model into an atrial-enriched fibroblast signaling network by adding atrial-specific components relevant to AF⁷. The present study builds on that atrial-enriched network and adopts Watts’ approach for representing the effects of sex hormones to study sex differences in AF.

3. E2 Inhibits Smad3 and Upstream Regulators and T Primarily via Smad3

The expanded model by Khorasani et al. successfully replicates the known anti-fibrotic effects of E2 and TS, validating it against experimental datasets with 81% accuracy in cardiac fibroblasts and 100% in atrial fibroblasts, which is on par or better than most logic-based models. The validated model provided new insights on pathways not previously shown experimentally. TS attenuated fibrosis primarily via Smad3 whereas E2 attenuated fibrosis through several nodes such as Smad3, AT1R, ROS, and JNK, with the ROS/JNK pair being a major contributor to E2’s anti-fibrotic effects. Knocking down ROS or JNK individually produced little effect, but their combined disruption suppressed all fibrotic markers. Thus, the model revealed that E2’s protection arises from the coordinated feedback between ROS and JNK, rather than a single node.

The authors extended this analysis by simulating young and aged female and male subjects, adjusting sex hormone levels to reflect physiological differences across the lifespan. Aged females resembled young males, consistent with the lower E2 levels after menopause. ROS/JNK knockdown blunted fibrosis in every group except young females, restoring the protection that E2 provide.

4. Limitations

The model contains several limitations that could be the subject of future work. As acknowledged by the authors, the network lacks the incorporation of other sex hormones such as progesterone. Including progesterone is important as its levels increase during pregnancy, a period where the risk of AF increases⁸. The model also does not include aromatase-dependent conversion of TS to E2. This conversion could allow TS to additionally engage with E2 related pathways, potentially broadening its protective mechanism. It is also worth noting that this model isolates the effects of sex hormones alone, but sex differences in AF could also arise from chromosomal and epigenetic regulation, independent of hormone levels. Disentangling chromosomal influences from hormonal ones will require models and data that go beyond circulating hormone levels.

The network integrates fibroblast data across species, ages, and cardiac chambers, and most pathways are derived from qualitative data, necessitating uniform reaction Hill parameters across the model. This is not unique to Khorasani’s model, but rather reflects the two-edged sword of logic-based models: on the one hand they facilitate integration of qualitative data in a framework that enables powerful quantitative predictions while, on the other hand, the lack of reaction-specific kinetics may affect the accuracy of these predictions.

These limitations could be addressed if sufficient experimental data would be available—a model is only as good as the data it is based on. However, experimental data with sex hormones as biological variable in (human) atrial fibroblasts remains scarce. Generating more experimental data on how E2 and TS act in atrial fibroblasts would sharpen computational models like the one presented by Khorasani et al. and move the field towards more accurate therapies.

5. From computational simulations to AF treatments

By integrating E2 and TS into an atrial-enriched fibroblast model, Khorasani et al. demonstrate the mechanistic basis of how sex hormones protect against atrial fibrosis. What is novel is the computational model’s identification of ROS and JNK as a combined target. E2 acts broadly through a coordinated ROS/JNK feedback converging on Smad3, while TS acts primarily via Smad3. The authors identify ROS/JNK as potential targets for therapeutic strategies, capable of restoring E2-like anti-fibrotic role.

Because fibrosis plays a key role in AF pathophysiology, treatments targeting fibrosis are of potential clinical interest. Many studies have shown the protective effect of E2. However, simply restoring E2 levels is often not the best treatment in men, as it has been associated with increased cardiovascular risk and gynecomastia⁹. The authors suggests that targeting ROS/JNK may capture E2’s role in modulating fibrosis. Because ROS/JNK knockdown replicated much of E2’s protection in males and aged females, while having little effect in young females, the therapeutic opportunity appears to be concentrated in the populations that need it. These findings suggest that the modulation of ROS/JNk may represent a strategy to mimic E2’s anti-fibrotic effects. Targeting ROS in AF itself is not new, as N-acetylcysteine (NAC) is already clinically used as an antioxidant to lower the incidence of postoperative AF¹⁰. However, as of writing, there are no approved drugs that target both ROS and JNK within AF. This is why computational models like Khorasani’s are useful: they point to a specific combination worth investigating experimentally. Turning predictions like these into validated therapies will require a continued collaboration between computational and experimental work, one where new experimental data refines computational models, and sharper computational models guide where the experimental data should come from next.