Evidence from several in vitro and animal studies indicated that an overactive immune response might lead to a storm of inflammatory cytokines, and some of these inflammatory cytokines directly affect the function of ion channels in cardiomyocytes. In this research, we mainly evaluated the effects of three cytokines (i.e., TNF-α, IL-1β, and IL-6) on cardiomyocytes and their possible proarrhythmic effects. It has been demonstrated that these cytokines can prolong the ventricular action potential duration (APD) by modulating several targets in cardiomyocytes, specifically the transient outward K⁺ channel (I _to), the rapid delayed-rectifier K⁺ current (I _Kr), and some targets involved in calcium handling. Focusing on acute inflammation, we screened out the experimental data based on the duration of the experimental treatment (less than 48 h), which are listed in Table 1, Table 2, and Table 3.

The rat ventricular cell model by Terkildsen et al. (referred to as the Terk model) (Terkildsen et al., 2008) and the human ventricular cell model by Ten Tusscher et al. (referred to as the TP06 model) (Ten Tusscher et al., 2006) were adopted in this study. Due to the lack of heterogeneity in the Terk model, AP heterogeneities, including transmural heterogeneity and interventricular heterogeneity, were incorporated according to our previous study (Bi et al., 2021) and experimental observations (Clark et al., 1993; Shimoni et al., 1995; Casis et al., 1998; MacDonell et al., 1998; Kaprielian et al., 1999; Ashamalla et al., 2001).

In the single-cell simulation, the rat model was paced with a series of 1000 stimuli with an amplitude of 6 pA/pF and a duration of 5.0 ms (80 pA/PF, 0.5 ms in the TP06 model) to reach the steady-state. To investigate the effects of a single cytokine and the combined influences of multiple cytokines on the cardiomyocytes, we adjusted the conductance of the related channel or ion flux of the related calcium handling process in the cell models according to the previous experimental recordings (Table 1, Table 2, Table 3). Note that the combined effects of the three cytokines were assumed to be an accumulation of each cytokine. The AP traces, APD₉₀, and current density traces of the different types of cells under various conditions were recorded for later analysis.

In addition, the data used in this study were obtained from bioexperiments in which the preparation concentrations of the cytokines were higher than the clinically measured cytokine levels in patients (Liu and Zhao, 1999; Monnerat et al., 2016; Liu et al., 2021). Therefore, we also considered another ‘mild’ type of inflammation (referred to as mild inflammation in this study) by halving the reported effects of cytokines as shown in Table 1, Table 2, Table 3.

The individual effect of each inflammatory cytokine and their combined effects (called “inflammation” in this study) on APs (1 Hz) are shown in Figure 1. First, for the rat cell model, it can be observed that except for IL-6 causing a negligible influence on EPI AP, all of these cytokines caused obvious AP prolongations in the ENDO/EPI cells. In terms of APD₉₀, both EPI and ENDO cells exhibited a significant increase compared with their control levels, which hinted at the presence of severe pathological repolarization heterogeneity between normal and inflammatory tissue. Next, in the human cell model, the simulation results (Figure 1B) showed that the APs in the IL-6 and TNF-α groups were prolonged in all three types of cells, but there was little change in AP in the IL-1β group. Moreover, as Figure 1Biv shows, the ΔAPD between MID and ENDO/EPI exhibited an obvious augmentation (from 97 to 150 ms) under inflammatory conditions, leading to a larger transmural repolarization heterogeneity compared with the control level.

In addition to the influences on AP, we also investigated the alteration of calcium handling using the TP06 model, as shown in Figure 2. In this regard, the reported reduced systolic [Ca²⁺]_i (Sugishita et al., 1999) and elevated diastolic [Ca²⁺]_i (London et al., 2003) under inflammatory conditions were successfully reproduced (Figure 2B). This observation might be attributed to the reduced SERCA activity and the increased SR Ca²⁺ leakage, which also caused a decreased SR Ca²⁺ content (Figure 2C). Moreover, the greatly decreased peak Ca²⁺ concentration in the cytoplasm exactly reflected the negative inotropic effect (Weisensee et al., 1993; Sugishita et al., 1999; Duncan et al., 2010) under inflammatory conditions.

The above simulation results were based on experimental data using high doses of cytokines. In this study, we also tested a type of mild inflammation by downregulating the reported effects in the TP06 model. The simulation results are shown in Figure 3. We can see that APD varied with the degree of inflammation, and in the case of mild inflammation, the ΔAPD between MID and ENDO/EPI cells showed a slight increase (from 97 to 119 ms) compared with the control group.

The other ionic current traces in the rat and human cell models under different cytokines can be found in Supplementary Figures S2,S3.

Unidirectional conduction blocks are an important pathological phenomenon in 1-D tissue, as the unidirectionally propagated excitation wave can evolve into reentrant spiral/scroll waves in 2-D slices and 3-D organs. The time window within which unidirectional conduction blocks occur, termed the vulnerable window, is a commonly used metric for measuring the temporal susceptibility of tissue to arrhythmias. In this section, we quantified the influence of inflammation on temporal vulnerability by measuring the VW across the 1-D transmural strand. Different degrees of inflammation, i.e., an extreme level of inflammation (e.g., sepsis) and a mild level of inflammation, were evaluated individually. The simulation results are shown in Figure 4. Specifically, Figure 4A shows different responses to the premature S2 stimulus, including bidirectional conduction (S2 too early), unidirectional conduction block (VW), and bidirectional conduction block (S2 too late). The distribution of the time window when unidirectional conduction blocks occurred (i.e., the distribution of VW) is plotted in Figure 4B. Both mild inflammation and inflammation delayed the occurrence of VW, and the average width of the VWs increased gradually from 7.4 to 8.8 and 10.0 ms depending on the degree of inflammation. As a wider VW signifies a higher chance of a unidirectional conduction block, the above simulation results implied an increased temporal susceptibility to reentry arrhythmias under inflammatory conditions.

Inflammation may initially occur in a specific area in the heart. This type of inflammation, termed local inflammation in this study, may result in the presence of pathological heterogeneity in ventricular tissue due to the prolonged APD of the affected cells, which predisposes to ventricular arrhythmias. In this section, we mimicked this condition by setting a local region of inflammation on the free wall of the left ventricle, and a typical S1-S2 protocol was used to evaluate the inducibility of reentry arrhythmias in this condition (see Methods for more details). The simulation results are shown in Figure 5 (rat) and Figure 6 (human).

First, for the control condition (Supplementary Video S1), it can be observed that, due to the limited size of rat hearts, the evoked unidirectional conduction was not able to turn back to form a functional reentry, and the subsequent two waves propagating along the ventricular wall collided and failed to form anatomical reentry (snapshots in Figure 5A). The whole process lasted only approximately 90 ms. For the human ventricular slice (Supplementary Video S2), although the evoked unidirectional conduction could turn back to form a functional reentry, such a process was unsustainable, and the spiral waves terminated shortly after the first cycle with a brief lifetime of approximately 350 ms (snapshots in Figure 6A).

Simulation results of local inflammatory conditions are shown in Figure 5B (rat) and Figure 6B (human). Due to the asynchronous repolarization caused by the prolonged APD of inflammatory cells, extra S2 stimulation applied to the border area would encounter the refractory tail on the inflammatory side. Therefore, S2 would generate a unidirectional conduction block, which in turn would evolve gradually into an anatomical spiral wave circling around the ventricular ring structure (Supplementary Video S3,S4). In brief, the pathological heterogeneity caused by inflammatory cytokines provided extra substrates for unidirectional conduction block and reentry arrhythmias.

Another type of inflammatory condition, as opposed to local inflammation, is global inflammation. It reflects a globally affected condition in which the whole ventricle is influenced by inflammatory cytokines. Compared to its counterpart, global inflammation does not create additional pathological heterogeneity; however, the prolonged wavelength may impair the tissue’s adaptability to fast heart rates. In this study, we mimicked the global inflammatory condition by setting all cells as inflammatory cells and tested the influences of global inflammation on the critical PCL using rat and human ventricular models (see Methods for more details). In the rat model, the measured critical PCL was 101 ms (9.9 Hz) in control conditions, while in global inflammation, it increased to 390 ms (2.56 Hz), and stimuli with cycle lengths below 390 ms led to complete repolarization failure (see Figure 7A or Supplementary Video S5). The critical PCL of 390 ms corresponds to a heart rate of 154 bpm, which is significantly lower than the physiological range of approximately 300–500 bpm in rats. Such simulation results showed a decreased adaptability of tissue to fast heart rates under global inflammation conditions and suggested a strong proarrhythmic effect. However, considering that the heart rate of rats is much faster than that of humans, this result might be species dependent. In addition, the atypical phenomenon of complete repolarization failure might also depend on the specific model (Supplementary Figure S6). To further clarify these questions, we performed parallel simulations using human tissue models. The simulation results showed that complete repolarization failure did not occur even at very high frequencies; however, there was still a critical PCL below which the 1:1 conduction could not be maintained and was replaced by 2:1 conduction (see Figure 7B or Supplementary Video S6). In this setting, the critical PCL was increased from 319 ms in the control condition to 380 ms in the global inflammatory condition. Notably, the BCL of 380 ms, which corresponds to a heart rate of approximately 157 bpm, was physiologically relevant and therefore might lead to arrhythmogenesis.

The generated pseudo-ECG of the human ventricular 2-D slice model under physiological and inflammatory conditions is shown in Figure 8. As some available studies concerning systemic inflammation have reported (Adlan et al., 2015; Lazzerini et al., 2020a; Armbruster et al., 2022), prolonged QT intervals were also reproduced in our simulation results.

In recent years, accumulating evidence has shown an association between systemic inflammation and cardiovascular disease. Inflammatory cytokines, a type of signaling molecule secreted from immune cells, were proven to be able to affect membrane ion channels and might therefore lead to ventricular arrhythmias. In this study, we selected three of these major cytokines and evaluated their proarrhythmic effects using a multiscale virtual heart. The main findings are as follows: 1) at the cellular level, inflammatory cytokines caused a prolongation of APD by affecting multiple ion channels, and heterogeneously prolonged APDs led to augmented transmural heterogeneities; 2) simulation results of the VW using the 1-D strand suggested that inflammation increased the temporal vulnerability to arrhythmias; 3) in the case of local inflammation, the repolarization of the inflammatory area was delayed due to the cytokine-induced APD prolongation, leading to the presence of pathological heterogeneities around the local inflammatory area. Such regional differences in repolarization provided extra substrates for the unidirectional conduction block and increased the chance of the development of anatomical reentry arrhythmias; 4) In the global inflammatory condition, the generated pseudo-ECG exhibited a prolonged QT interval that was in accordance with the clinical observations. Furthermore, the globally prolonged APD impaired the tissue adaptability to high frequencies and caused 2:1 conduction at physiologically relevant heart rates.

APD prolongation has been observed in many pathological conditions and has been shown to be proarrhythmic (Antoniou et al., 2017; Dietrichs et al., 2019; Shimaoka et al., 2020). For example, APD prolongation and changes in electrophysiological characteristics caused by the downregulation of outward potassium currents, alterations of calcium channel kinetics and increases in late sodium currents create a substrate for ventricular arrhythmias in the case of heart failure (Zhang et al., 2018). A study on hypokalemia showed that APD prolongation could predispose patients to early afterdepolarizations, which in turn act as triggers for ventricular arrhythmias (Tse et al., 2021). APD prolongation is also the main factor responsible for the proarrhythmic influences of drugs with cardiotoxicity. A prolonged APD and, consequently, a prolonged QT interval, are considered important biomarkers of drug cardiotoxicity in drug discovery (Schrickel et al., 2006; Hondeghem et al., 2011). For example, the antihistamine drug terfenadine, which was previously used for the treatment of allergic conditions, was proven to be able to prolong APD due to its effects of blocking hERG currents (I _Kr), causing QT prolongation and torsade de pointes (TdP). Despite the tight association between APD/QT prolongation and ventricular arrhythmia, prolonged QT does not always lead to arrhythmogenesis events. Some new evaluation criteria, such as TRIaD (Triangulation, Reverse use dependence, Instability and Dispersion), have been suggested to identify false-positive cases. On the other hand, it should also be noted here that APD prolongation can also exert antiarrhythmic effects by extending the effective refractory period, which is also the major pharmaceutical mechanism of class III antiarrhythmic drugs. In particular, homogeneous APD prolongation in the absence of early afterdepolarizations (EADs) commonly exerts antiarrhythmic effects, whereas EADs or repolarization heterogeneities induced by inhibition of repolarization are proarrhythmic. For the case in this study, the simulation results demonstrated a double-sided nature, especially in a condition of global inflammation: the prolonged refractory period in ventricles tends to be antiarrhythmic, but it also reduces the adaptability of the tissue to high heart rates. For the local inflammatory condition, although we did not observe EADs, the presence of both pathological heterogeneities and augmented transmural heterogeneity increased the vulnerability to arrhythmogenesis in terms of both spatial and temporal aspects.

Accumulating studies have demonstrated that augmented electrophysiological heterogeneity provides proarrhythmic substrates (Laurita and Rosenbaum, 2000; Pak et al., 2004; Antzelevitch, 2007; Boukens et al., 2009). Intrinsic heterogeneity within normal hearts contributes to the development of arrhythmias under certain conditions (Bishop et al., 2013). Premature beats in the heterogeneous area that occurred during a certain time duration (i.e., VW) would lead to unidirectional conduction block and reentry arrhythmias. In most cases, the VW is rather small; however, it can be significantly enlarged in some pathological conditions due to augmented regional heterogeneity, such as ischemia (De Bakker et al., 1988; Yuuki et al., 2004) and heart failure (Roden, 2003; Coronel et al., 2013). For the local inflammatory condition, our simulation results suggested that the repolarization of the tissue developing inflammation was significantly delayed, which in turn contributed to an apparent electrical heterogeneity around the local area. Such regional dispersion of repolarization, as an arrhythmogenic substrate, will be further amplified in higher but physiologically relevant pacing frequencies and eventually lead to arrhythmia. Regarding interspecies differences, the simulation results suggested that inflammation weakened the repolarization ability of both rat and human myocytes, but there were obvious interspecies variances behind it, as repolarization is notably different in rodents compared to large mammals. In particular, the cytokine-induced APD prolongation in rat myocytes was mainly attributed to the decreased I _to; in contrast, the APD prolongation in human ventricular cells resulted mainly from the reduced I _Kr. It has been demonstrated that I _to is the major repolarization current in rat myocytes (Zhao et al., 2012), but for humans, I _to is only one of various repolarizing currents and is mainly involved in repolarizing phase 1 of cardiac AP. In comparison, hERG (the gene that codes for the alpha subunit of I _Kr) is abundantly expressed in human ventricles, and I _Kr plays a critical role in repolarizing during cardiac AP (Ledford et al., 2022; Cheng and Kodama, 2004). This interspecies variation has also been widely observed in other studies. For instance, when used to treat carbon monoxide (CO)-induced arrhythmias, ranolazine was shown to be effective in inhibiting CO-induced EAD in rat cells (Dallas et al., 2012), but it exacerbated EADs and even caused oscillation in guinea pigs (which have wide APs similar to those in humans). In a recent study (Morotti et al., 2021), Morotti et al. examined the influences of interspecies differences on animal experimentation and drug efficacy assessment, and they created cross-species translators of electrophysiological responses to translate the drug-induced effects experimentally observed in myocytes from animal models to predict the effects that these perturbations would cause in humans. In summary, although our simulations demonstrated that inflammation caused similar reentry arrhythmias in rats and humans, the underlying mechanisms were different among various species and this needs to be taken into account, especially when translating experimental findings regarding drug efficacy and safety from animal models to clinical use.

It should be noted that the combined effect of three cytokines was assumed to be an additive effect of each cytokine in the present work. Kumar et al. (Kumar et al., 1996) reported a synergistic inhibitory effect of TNF-α and IL-1β on ventricular contractility, where the concentration of each cytokine under conditions of a combination was much lower than those acting alone. Such synergistic effects and the potential antagonistic effects among cytokines were not considered in this study.

Although the present study followed the experimental setup in the original reports exactly, the concentrations of inflammatory cytokines used in previous experiments were generally higher than those observed under pathophysiological conditions. Specifically, Liu et al. reported a median IL-6 serum level of 4809 pg/ml in a nonsurvivor group of patients with sepsis (Liu et al., 2021). Liu et al. reported that the pathophysiological level of TNF-alpha was approximately 0.15 ng/ml in the serum of heart failure patients (Liu and Zhao, 1999). Monnerat et al. reported a serum IL-1β level of 60 pg/ml in diabetic mice (Monnerat et al., 2016). Therefore, our simulation results reflected an extreme condition of inflammation (i.e., sepsis). To account for this, we have considered another ‘mild’ inflammatory condition by halving the concentrations of the inflammatory cytokines; however, caution should still be taken when translating these findings to clinical use.

The effects of inflammatory cytokines on ion channels were assumed to be consistent by setting specific constant change ratios to the conductance of the involved ion channels. However, inflammation is a dynamic process with complicated mechanisms. For example, there have been studies showing that inflammatory cytokines could increase sympathetic activity to inhibit cytokine production; however, it also deserves to be noted that hyperactive sympathetic nerves could directly influence the function of ion channels by phosphorylation in myocardial cells and may induce arrhythmias (Lazzerini et al., 2017). The above dynamic process, as a kind of negative feedback mechanism, could change the concentrations of inflammatory cytokines in serum, which was not incorporated in our simulations. In addition, inflammatory cytokines could also affect the electrophysiological function via complicated indirect pathways in addition to the introduced direct modulations of ion channels and calcium handling. These indirect pathways include, but are not limited to, the following two aspects: 1) inflammatory cytokines could cause chronic remodeling and myocardial fibrosis, thus increasing the susceptibility of cardiac tissue to arrhythmias in chronic heart failure patients (Dick and Epelman, 2016); 2) current evidence suggests that TNF-α may promote the formation of atherosclerotic plaques by upregulating the expression of multiple protein molecules (e.g., adhesion molecule-1) in the vascular wall (Ohta et al., 2005), which would exacerbate the ischemic condition of hearts and lead to ischemic-related arrhythmias. As we mentioned above, inflammation has complex mechanisms and effects that require further simulation studies. Above all, these indirect pathways play critical roles in inflammation-mediated arrhythmias but they were not investigated in this study.

Limited to the sources of the experimental data, the proarrhythmic mechanisms investigated in this study focused mainly on ventricular arrhythmia. In fact, accumulating evidence has shown that there is a strong link between inflammation and postoperative atrial fibrillation. Maesen et al. reported an overlapping time course of atrial fibrillation occurrence after cardiac surgery and the activation of the complement system with the release of proinflammatory cytokines, suggesting potential roles of inflammation in triggering postoperative atrial fibrillation (POAF) (Maesen et al., 2012). Heijman et al. observed that postoperative inflammation along with preexisting Ca²⁺-handling abnormalities contributed to the formation of DADs and thus led to POAF (Heijman et al., 2020). The above postoperative atrial fibrillation, as a widely accepted form of inflammation-induced arrhythmia, warrants further research.