Acute ischemia/reperfusion increases AF susceptibility. The main purpose of this study was to model the ischemic process of the donor heart and to determine the effect of different durations of ischemia followed by reperfusion on the susceptibility to AF. Langendorff perfusion of isolated rat hearts was used followed by endocardial optical mapping of isolated atrial preparations, allowing us to investigate the effects of ischemia/reperfusion (I/R) injury ex vivo on AF susceptibility (Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.185961DS1). In brief, after stabilizing the baseline and administering pretreatment, global no-flow ischemia was induced for 0 minute (nonischemia, IS-0), 10 minutes (IS-10), or 25 minutes (IS-25), followed by 30 minutes of reperfusion (Figure 1A). The voltage-sensitive dye RH237 was utilized to observe changes in action potentials. The atria were then dissected, transferred to a perfusion chamber with the endocardial surface up, and continuously superfused with Tyrode’s solution. Programmed electrical stimulation was performed to assess tissue electrical conduction characteristics. Burst pacing was repeated 4 times (n = 5) to determine AF inducibility. An episode of flutter or fibrillation lasting over 2 seconds was considered a positive outcome. Subsequent bursts were delivered 1 minute after an AF episode terminated or 1 minute after a noninducing burst.

Acute atrial I/R–induced AF and SAN dysfunction. (A) Experimental protocol. Stab, stabilization; IS-0, no ischemia without insulin; IS-10, ischemia for 10 minutes without insulin; IS-25, ischemia for 25 minutes without insulin. (B) Atrial ECG of a short episode of AF induced by a 2-second 50 Hz burst pacing. (C) Optical mapping of action potential of an episode of AF. (D) AF inducibility. (E) RAF duration. Ischemia for 25 minutes significantly increased the susceptibility to RAF and cumulative duration of RAF. (F) The right atrium is 8 times more susceptible to AF than the left atrium after 25 minutes of ischemia. (G) We found 60% RAF episodes were due to triggered activities, half of which originated from and/or near the SAN area. (H) Activation map of atria; arrow indicates the identification of SAN. (I) Action potential of SAN (near the SAN area) and sinoatrial node recovery time (SANRT). (J) Impact of different ischemia duration on SANRT. Data in all panels except for E were presented as mean ± SEM; n = 6 (biological repeats) each group; statistical analysis was performed by 2-way (for D) or 1-way (for J) ANOVA, respectively, with Tukey’s multiple-comparison test. Data in E were presented as median [25th to 75th percentile] and analyzed by Kruskal-Wallis test; n = 5, 8, 25 (technical repeats) in IS-0, IS-10, and IS-25, respectively.

Representative AF episodes recorded by electrocardiography (ECG) and optical mapping are shown in Figure 1, B and C, respectively. The inducibility of AF via right atrial fibrillation (RAF) bursts among atria from hearts subjected to 25 minutes of ischemia was significantly higher than among atria subjected to 10 minutes of ischemia or no ischemia (71.67% ± 9.80% vs. 10.0% ± 3.65% in IS-10, or 3.33% ± 2.11% in IS-0, both P < 0.01, Figure 1D). There was no significant difference in RAF inducibility between nonischemic atria and atria undergoing 10 minutes of ischemia, and there was no significant difference in AF inducibility via left atrial burst stimulation (left atrial fibrillation, LAF) among all groups (Figure 1D). Following 25 minutes of ischemia, AF inducibility was markedly lower in the left versus right atria (8.33% ± 3.07% vs. 71.67% ± 9.80%, P = 0.0024, Figure 1D). Supplemental Table 1 shows occurrence of AF episodes in each animal. Although we tested the inducibility of RAF and LAF separately and AF induced from either site always propagated to the opposite site, we only counted the occurrence of AF at the atria with pacing. These data suggest that the right atrium is more vulnerable to atrial I/R–induced AF than the left atrium.

AF is considered positive when its duration lasts longer than 2 seconds to avoid false-positive results. The duration of AF was compared to evaluate the severity of AF. Ischemia for 25 minutes significantly increased the cumulative duration of RAF (178 [81 to 380.5] seconds; median [25% to 75% interquartile range]), compared with 10 minutes of ischemia (17.5 [8.25 to 84] seconds; P = 0.0136) or no ischemia (8 [5.25 to 13.75] seconds; P = 0.0038; Figure 1E). We found that the right atrium was 8 times more vulnerable than the left atrium following 25 minutes of ischemia (Figure 1F). Among the RAF episodes, 60% were identified as triggered activity–related AF. Representative images of triggered activities and reentrant AF are shown in Supplemental Figures 2 and 3. Of these, half originated from the anatomic region delineated by the sinoatrial node (SAN), superior vena cavae, and the coronary sinus (Figure 1G).

The SAN region was verified at baseline before burst pacing, and the sinoatrial node recovery time (SANRT) was collected (Figure 1, H and I). After 25 minutes of ischemia, there was a significant increase in SANRT compared with nonischemic conditions (1,129 ± 111.5 ms vs. 293.4 ± 38.7 ms, P = 5.39 × 10–6) and after 10 minutes of ischemia (1,129 ± 111.5 ms vs. 578.9 ± 68.51 ms, P = 0.0005) (Figure 1, I and J). Compared with nonischemic conditions, there was prolongation of SANRT after 10 minutes of ischemia, but the change did not reach statistical significance (P = 0.0534). These findings suggest that acute atrial I/R–induced AF adversely affects the SAN function and that ischemia duration plays a crucial role for determining AF susceptibility.

Reentrant circuits were observed in approximately 40% of cases. An illustrative activation mapping of an episode can be found in Supplemental Figure 3, A and B. This particular episode of AF was identified as a macro reentrant circuit initially resembling fast atrial arrhythmia but later transitioning into fibrillation as the action potential patterns became more erratic (Supplemental Figure 3, C and D). As depicted in Supplemental Figure 4, A–C, though ECG from the entire atria shows the episode of AF (Supplemental Figure 4A) is chaotic, the action potential of a single pixel is similar (Supplemental Figure 4B). However, the mapping revealed that each action potential followed a distinct conduction pathway (Supplemental Figure 4C), ruling out AF with a fixed origin and suggesting a combination of triggered activity, reentrant circuits, and heterogeneous tissue properties. This highlights the complexity of the mechanism of acute atrial I/R–induced reentrant AF, underscoring the substantial impact of acute atrial I/R–induced tissue damage on increasing electrical conduction variability.

Insulin prevents I/R-induced AF. We investigated whether insulin modulates AF inducibility by I/R. The ex vivo study enabled us to test this hypothesis using a higher insulin dose. Figure 2, A and B, show that insulin (5 mU/mL) significantly reduced RAF inducibility compared with nontreated control (10% ± 4.47% vs. 71.67% ± 13.76%, P = 0.0038) and shortened RAF duration (196.5 [112.3 to 282.3] seconds vs. 12 [8.75 to 28] seconds, P = 0.0003, Figure 2C). Supplemental Table 2 summarizes the occurrence of AF episodes in each animal. To our knowledge, this is the first report of insulin providing protection against acute atrial I/R–induced AF.

Insulin prevents I/R-induced RAF. (A) Experimental groups. IS-25, ischemia for 25 minutes without insulin; IS-25ins, ischemia for 25 minutes with insulin; 11.3GLU, treatment with high concentration of glucose (11.2 mM) without insulin; PRV, treatment with pyruvate without insulin; PRVins, treatment with pyruvate plus insulin; Insulin-reperfusion, insulin treatment during reperfusion only. (B) Inducibility of RAF. (C) RAF duration. (D) SANRT. (E and F) Effective refractory period (ERP). Horizontal bars represent individual comparisons (P < 0.05). (G) Pacing threshold. Data in all panels except for C were presented as mean ± SEM; n = 6 (biological repeats) per group; statistical analysis was performed by 1-way ANOVA with Tukey’s test. Data in C were presented as median [25 to 75 percentile]. n = 14, 11, 14, 6, 7, and 7 (technical repeats), respectively, in IS-25, IS-25ins, 11.2GLU, PRV, PRVins, and Insulin-reperfusion; statistical analysis was performed by Kruskal-Wallis test with Dunn’s multiple-comparison test.

Given that insulin can directly facilitate the transport of glucose into cardiomyocytes via glucose transporters, leading to increased glucose uptake, we developed a subhypothesis that the protective effect is dependent on increased glucose uptake. Therefore, we compared the effects of higher extracellular glucose levels, replacing glucose with equimolar pyruvate, and using 2-deoxy-d-glucose (2DG; an inhibitor of glucose metabolism, data not shown), as shown in Figure 2A. Elevated glucose did not reduce susceptibility to AF compared with regular glucose (P = 0.9994, Figure 2B). Substituting glucose with pyruvate (5.6 mM/L) in the presence or absence of insulin showed that insulin with pyruvate prevented the onset of AF compared with insulin alone (P = 0.0268), suggesting that insulin’s protection against AF is not dependent on glycolysis. However, using insulin with 2DG did not show a substantial benefit, likely due to cell energy deprivation and resulting dysfunction, since glycolysis was the primary energy source in this experimental setup. Insulin with glucose or pyruvate prevented the prolongation of SANRT (Figure 2D). There were no significant differences in the effect on the refractory period among conditions with 25 minutes of ischemia (Figure 2, E and F). With insulin present, the pacing thresholds were lower than in the 25-minute ischemia condition and similar to the 10-minute ischemia condition, but this effect was not observed when insulin was administered only during reperfusion (Figure 2G). These findings suggest that the protective effect of insulin against AF is not reliant on extracellular glucose and is effective only when administered prior to ischemia.

Insulin prevented I/R-induced electrical heterogeneity. S1S1 pacing at various cycle lengths including 150, 130, 120, 110, 100, 90, 80, and 70 ms was used to assess action potential duration (APD) and conduction velocity. As depicted in Figure 3, A–E, optical mapping demonstrated that after 25 minutes of ischemia, there was a decrease in APD at 50% and 90% repolarization (APD50 and APD90), a reduction in conduction velocity at different pacing cycle lengths (PCLs), and an increase in conduction heterogeneity. However, insulin prevented the decrease in APD and slowing of conduction velocity, resulting in improved electrical conduction homogeneity compared with the control group subjected to 25 minutes of ischemia. Separate experiments were then conducted using Rhod-2 AM dye to gather calcium transient signals. Tissues exposed to 25 minutes of ischemia without insulin showed more calcium alterations and changes in waveforms (Figure 4, A–D). Moreover, tissues exposed to 25 minutes of ischemia without insulin exhibited longer calcium transients (Figure 4, E–G).

Insulin attenuates I/R-induced atrial electrical remodeling. (A and B) APD at 50% or 90% repolarization (APD50 or APD90). PCL, pacing cycling length. (C) Conduction velocity. (D and E) Conduction heterogeneity (D) and activation map (E) of the right atrial tissue at 130 ms PCL. Data in all panels were presented as mean ± SEM; n = 6 (biological repeats) each group; statistical analysis was performed by 2-way ANOVA (for A–C) or 1-way ANOVA (for D) with Tukey’s multiple-comparison test. †P < 0.05, ‡P < 0.01 compared with IS-25ins.

Insulin improves calcium transient in ischemic atrial tissues. (A–D) Representative recordings of calcium transient alterans (CaTA) (A) and waveform changes (C) paced at a cycling length of 130 ms and the quantification of CaTA (B) and waveform changes (D). The arrows indicate abnormal calcium transition. The percentage in B represents the percentage change of the later (dashed line b) from the earlier (dashed line a) calcium transient amplitude. The percentage in D represents the percentage waveform changes after the S1S1 pacing. (E) Typical calcium transients. (F and G) Calcium transient duration at 80% recovery (CaTD80) and amplitude. Data were presented as mean ± SEM; n = 6 (biological repeats) each group. Statistical analysis was performed by 2-tailed Student’s t test (for B and D) or 2-way ANOVA (for F and G) with Tukey’s multiple-comparison test.

We analyzed the primary regions of an episode of AF. Primary regions are the areas of the heart that are the sources or focal points of an arrhythmic event, as illustrated in Figure 5, A and B. To compare the distribution of the primary regions, these points were plotted for Poincaré plot analysis. The standard deviations SD1 and SD2 were calculated from these plots to provide insight into the variability of the leading regions. The mean SD (SDmean) was then determined to explain the distribution of the primary regions (Figure 5C). The results showed that ischemia for 25 minutes increased the SDmean value compared with the nonischemic control (P = 0.0379, Figure 5D) or those with 10 minutes of ischemia (P = 0.0012). A significantly lower SDmean was observed in the insulin-treated group (P = 0.0001). These findings quantified the distribution of the primary regions in an episode of AF. High dispersion suggests that the AF episodes originate from multiple and varied locations, while low dispersion indicates that the arrhythmic events are more localized or confined to specific areas. Therefore, these results suggest that electrical heterogeneity within the atrial tissue is severe in those under 25 minutes of ischemia, whereas insulin-treated tissue shows larger conduction homogeneity. This increased heterogeneity could contribute to the increased susceptibility and complexity of AF under prolonged ischemic conditions.

Insulin reduces the dispersion of leading regions. (A) Examples of AF episode’s action potentials (blue) and simultaneous ECG recording (green). (B) Typical leading region mappings of different conditions. (C) Poincaré plot of the leading regions. (D) Dispersion with the use of SDmean value of SD1 and SD2 of the Poincaré plot. Data were presented as mean ± SEM; n = 3, 6, 6, and 6 (biological repeats), respectively, in IS-0, IS-10, IS-25, and IS-25ins; statistical analysis was performed by 1-way ANOVA with Tukey’s multiple-comparison test.

Insulin prevents I/R-induced SAN dysfunction. Understanding spontaneous AF during and after ischemia is crucial. To quantify the severity of spontaneous AF, an arrhythmia scoring system was developed: 0 for none, 1 for an episode lasting less than 2 seconds, 2 for greater than or equal to 2 but less than 15 seconds, 3 for greater than or equal to 15 but less than 30 seconds, 4 for greater than or equal to 30 but less than 60 seconds, 5 for greater than or equal to 60 but less than 120 seconds, and 6 for greater than or equal to 120 seconds. Additionally, the stop time of sinus rhythm (SRST) after ischemia and recovery time of sinus rhythm (SRRT) after reperfusion were measured to evaluate the function of the SAN.

Figure 6A shows a representative baseline electrocardiogram with clear and equal P waves, while Figure 6B shows a representative spontaneous AF where P waves disappeared and were replaced by f waves. A higher score was only found in the IS-25 group compared with IS-0, IS-10, and IS-25ins (P = 0.0004, 0.0033, and 0.0077, respectively, Figure 6C), suggesting that 25 minutes of ischemia increased susceptibility to spontaneous AF, but the use of insulin was protective. As shown in Figure 6, D and E, representative electrocardiograms display changes from baseline to cardiac arrest and recovery after ischemia, which include the (i) baseline, (ii) J point depression at the first minute of ischemia, (iii) J point and ST segment elevation at the third minute of ischemia, (iv) atrioventricular dissociation at the sixth minute of ischemia, (v) ventricular asystole before reperfusion, and (vi) sinus rhythm recovery after a certain period. There were no significant differences in SRST among ischemia conditions with or without insulin intervention. However, SRRT in 25 minutes of ischemia was significantly prolonged compared with 10 minutes of ischemia (P = 0.0039, Figure 6F), while with insulin, SRRT was shortened (P = 0.0049, Figure 6, F and G). These data support the conclusion that insulin provides protection against I/R-induced SAN dysfunction, aligning with the findings of SANRT after burst pacing.

Insulin reduces the susceptibility to spontaneous AF and SRRT after reperfusion. (A) Sinus rhythm. P waves were clear and identifiable. (B) An episode of AF. P waves were replaced by f waves, which suggested an episode of AF. (C) Scores of the spontaneous AF during I/R. (D and E) ECG changes during perfusion. ECG at baseline (i), first minute of ischemia (ii), third minute of ischemia (iii), sixth minute of ischemia (iv), before reperfusion (v), and after reperfusion (vi). (F) SRST. (G) SRRT. Data were presented as mean ± SEM; n = 6, 6, 9, and 9 (biological repeats), respectively, in IS-0, IS-10, IS-25, and IS-25ins; statistical analysis was performed by 1-way ANOVA with Tukey’s multiple-comparison test.

Insulin prevents I/R-induced change of transcriptome in the right atrium. To investigate the mechanisms of POAF and the protective effects of insulin, SAN tissue from various groups was collected for bulk RNA sequencing. In the 25 minutes of ischemia group, 1,182 genes were upregulated and 726 genes downregulated (FDR < 0.05) compared with the nonischemia group. The IS-25ins group showed fewer differentially expressed genes (DEGs) than the IS-25 group, with 966 genes upregulated and 670 genes downregulated, respectively. Only 461 DEGs were identified when comparing IS-25ins with IS-25 (Figure 7A). Principal component analysis (PCA) demonstrated distinct clustering of the samples in the comparison between nonischemia and IS-25 groups, as well as between IS-25 and IS-25ins groups (Figure 7, B and C). The top 5 altered Gene Ontology (GO) pathways in the ischemia for 25 minutes versus nonischemia comparison included positive regulation of cell adhesion, mononuclear cell differentiation, myeloid cell differentiation, and response to hypoxia (Figure 7D). In the comparison between IS-25ins and IS-25, the predominant pathways were positive regulation of cell adhesion, rhythmic process, regulation of neurogenesis, cell-substrate adhesion, and muscle organ development (Figure 7E). I/R induced substantial gene expression changes, further exacerbated by insulin treatment in the positive regulation of cell adhesion GO term (Supplemental Figure 5). Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis revealed alterations in the MAPK pathway (Supplemental Figure 6) and PI3K/AKT pathway (Supplemental Figure 7) in the IS-25 group compared with IS-0, with insulin partially reversing these changes, highlighting its role in modulating cell apoptosis and stress response. Additionally, cluster analysis categorized DEGs across the 3 comparisons into 16 distinct clusters (Figure 7F). Insulin treatment partially reversed gene expression changes in clusters 10–15 induced by I/R. In clusters 1–9 and 16, the gene expression patterns between the IS-25 and IS-25ins groups were similar (Supplemental Figure 8). Notably, clusters 11 and 12 were enriched in the circadian rhythm pathway, and cluster 15 was enriched in the ferroptosis pathway (Supplemental Data 1).

Gene expression and pathway analysis of right atrial myocardium. (A) Bar graph showing the number of DEGs between IS-0, IS-25, and IS-25ins groups. (B and C) Principal component analysis of gene expression across different groups (IS-0 vs. IS-25 in panel B and IS-25 vs. IS-25ins in panel C). (D and E) Dot plot of GO enrichment analysis across different groups (IS-0 vs. IS-25 in panel D and IS-25 vs. IS-25ins in panel E). (F) Cluster analysis of DEGs across the 3 groups. n = 6, 5, 5 (biological repeats) in IS-0, IS-25, and IS-25ins, respectively.

Insulin does not substantially change the abundance of intermediates in the glucose metabolism pathway. Tissues from right atrial myocardium of the IS-0, IS-25, and IS-25ins groups were collected for ion chromatography mass spectrometry (IC-MS), covering glycolysis, the TCA cycle, the pentose phosphate pathway, nucleotide metabolism, and other metabolic pathways. A total of 119 metabolites were detected and passed quality control. PCA revealed distinct clustering of the 3 groups (Supplemental Figure 9A). Forty-seven metabolites were significantly different between the IS-0 and IS-25 groups, with adjusted P values below 0.05; 16 were downregulated and 31 were upregulated in the IS-25 group (Supplemental Figure 9B). Insulin pre-perfusion in the IS-25ins group led to 19 downregulated and 11 upregulated metabolites compared with IS-25 (Supplemental Figure 9C). Pathway analysis showed that ischemia for 25 minutes followed by 30 minutes of reperfusion affected several pathways, including sucrose metabolism, fructose and mannose metabolism, alanine, aspartate and glutamate metabolism, the TCA cycle, glycolysis, pentose phosphate pathway, and others (Supplemental Figure 9D). Insulin pre-perfusion affected sucrose metabolism, taurine and hypotaurine metabolism, pentose phosphate pathway, fructose and mannose metabolism, alanine, aspartate and glutamate metabolism, amino sugar and nucleotide sugar metabolism, and glycolysis compared with IS-25 (Supplemental Figure 9E).

We analyzed the abundance of the metabolites in glucose metabolism to investigate if insulin affects glucose metabolism in the atrial myocardium. First, I/R in the IS-25 group resulted in decreased levels of glucose, glucose 6-phosphate, and fructose 6-phosphate and increased levels of glycerol 3-phosphate, 2-phosphoglycerate, and phosphoenolpyruvate compared with the IS-0 group. Insulin pretreatment in the IS-25ins group led to increased levels of glucose 1-phosphate, glucose 1,6-bisphosphate, glucose 6-phosphate, and fructose 6-phosphate compared with the IS-0 group, as well as increased levels of glycerol 3-phosphate compared with IS-25 (Supplemental Figure 10). These data implicate reduced glucose utilization in ischemic right atrial myocardium and increased glucose uptake induced by insulin. Second, neither I/R nor insulin pretreatment resulted in consistent changes of metabolites in TCA cycle and the levels of pyruvate and lactate (Supplemental Figure 10). Third, insulin treatment did not change ATP production in the ischemic atrial tissue (Supplemental Figure 11, A–C). Therefore, it is unlikely that altered glucose metabolism substantially contributed to insulin-mediated cardioprotection against ischemia-induced AF in the current study.

Insulin reduced both cardiomyocyte and noncardiomyocyte apoptosis. It was reported that cardiomyocyte apoptosis contributes to the pathogenesis of AF in patients and preclinical animal models (25, 26). Therefore, we measured cell apoptosis in right atrial heart tissue sections using terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining. Our data showed that ischemia for 25 minutes significantly increased apoptosis of both cardiomyocytes and noncardiomyocytes compared with nonischemic conditions, and treatment with insulin markedly reduced apoptosis of both cell types (Figure 8, A and B). There was no difference in the morphology of cardiomyocytes, sarcomere structure, and interstitial fibrosis between nonischemic and ischemic right atria regardless of insulin treatment, as depicted in Figure 8, C and D.

Insulin reduces cardiomyocyte apoptosis. (A) Cell apoptosis is measured by TUNEL staining. Cell nuclei were counterstained by DAPI. Bar = 100 μm. (B) Quantification of TUNEL positively staining nuclei of cardiomyocytes and nonmyocytes, respectively. The numbers of apoptotic cell nuclei were normalized to the total number of nuclei for cardiomyocytes and nonmyocytes, respectively. Data were represented as mean ± SEM. n = 5 (biological repeats) in each group. One-way ANOVA with Tukey’s multiple-comparison test. (C) H&E was performed to observe cardiac inflammation and gross morphology. Bar = 100 μm. Fast green/Sirius red staining was performed to determine the interstitial fibrosis. Bar = 100 μm. Immunostaining using antibodies against cardiac troponin T (cTnT) was performed to visualize cardiomyocyte morphology. Bar = 25 μm. (D) Quantification of interstitial fibrosis. Data were represented as mean ± SEM. n = 5 (biological repeats) in each group. One-way ANOVA with Tukey’s multiple-comparison test.