Distinct circadian mechanisms govern cardiac rhythms and susceptibility to arrhythmia

Abstract

Electrical activity in the heart exhibits 24-hour rhythmicity, and potentially fatal arrhythmias are more likely to occur at specific times of day. Here, we demonstrate that circadian clocks within the brain and heart set daily rhythms in sinoatrial (SA) and atrioventricular (AV) node activity, and impose a time-of-day dependent susceptibility to ventricular arrhythmia. Critically, the balance of circadian inputs from the autonomic nervous system and cardiomyocyte clock to the SA and AV nodes differ, and this renders the cardiac conduction system sensitive to decoupling during abrupt shifts in behavioural routine and sleep-wake timing. Our findings reveal a functional segregation of circadian control across the heart's conduction system and inherent susceptibility to arrhythmia.

Fig. 1. Longitudinal ECG reveals differential circadian regulation of sinoatrial (SA) and atrioventricular (AV) node function and decoupling by mistimed sleep.

A. Schematic of 4-day laboratory session, with total sleep deprivation (TSD) and recovery nap on day 3 (n = 14 individuals). B–F HR (B), HRV (C), and z-scored ECG parameter profiles under baseline (black) and TSD (orange) days (shading indicates + /−SEM), highlighting the profound impact of mistimed sleep on RR (D) and QT (E), but not PR_seg (F). HR has been derived from the RR interval and is shown for clarity. At baseline, all parameters were rhythmic based on cosinor analysis (P < 0.001). Individual traces were excluded from waveform analysis where data coverage fell <70% of the 5-min time bins; n (baseline/TSD) = 13/14 (B), 12/14 (C), 13/14 (D), 11/12 (E), 12/14 (F). Mean ECG parameters were quantified across a 4 h mid-night and mid-day (equivalent to the nap window on day 3) analysis windows on the baseline and TSD days (two-way RM ANOVA/Mixed model; n = 14 (B, C, D, F), 12 (E)). G, H Under baseline conditions, LOWESS fit of ECG profiles (G) and cross-correlation (H; yellow: RR vs RR, blue: QT vs RR, orange: PR_seg vs RR) revealed a significant phase delay in PR_seg rhythm relative to that of RR (Gaussian fit with one-sample T test; n = 13, 11, 12, respectively). I Acute changes in RR were mirrored by a concordant change in QT, but not PR_seg duration (ΔRR reflects z-scored difference in RR between sequential 5-min analysis bins; Δparam reflects the concurrent change in QT or PR). All data presented as group mean ± SEM; ns P > 0.05, **P < 0.01, ***P < 0.001. bpm = beats per minute; std = standard deviation. See Supplementary Fig. S2 for additional information; source data and statistical details are provided as a Source Data File.

Fig. 2. ECG response under simulated night-shift and constant routine.

A Schematic of 6-day laboratory session with simulated day-shift (left) and night-shift (right) routines (n = 7 individuals/group). B, C ECG parameters recorded during mid-day (pink) and mid-night (green) showed a rapid reversal of RR and QT interval rhythms, but not that of PR_seg in response to the switch to night-shift behavioral routine. Two-way ANOVA with repeated measures, colored asterisks indicate the difference from day 1. D Hourly binned z-scored group data and sinusoid fits from day- (orange) and night-shift (blue) conditions. Data presented as group mean ± SEM. E, F Timing (acrophase) of individual rhythm peaks relative to lights on (i.e., start of constant routine) (E) or external clock time (F). Asterisks indicate differences between groups (Watson–Williams test). *P < 0.05, **P < 0.01, ***P < 0.001. Source data and statistical details are provided as a Source Data File.

Fig. 3. Longitudinal ECG recording reveals differential circadian regulation of SA and AV node function in mice.

A Rhythms in locomotor activity (LA, arbitrary units) and ECG-derived heart rate (HR), HRV, RR, QT, and PR_seg parameters in mice (based on 5 days of recording; shading reflects + /− SEM; x axis black bar indicates dark phase; n = 10 mice). B–E Impact of LA and HR on ECG parameters. B Decreased LA during mid-night siesta (marked by an arrow in A) was accompanied by a significant decrease in RR and QT intervals, but not in PR_seg duration (2-h siesta period (black) vs preceding 2 h of activity (blue); two-way RM ANOVA). C Periods of LA followed by >45 min of complete inactivity were isolated and aligned to the cessation of activity (time 0). RR (yellow), QT (blue), and to a lesser extent PR_seg (orange) showed a significant response to the activity which decreased over subsequent inactivity (two-way RM ANOVA, Dunnet’s post hoc, difference from t = 40). D Transient bouts of LA (preceded and followed by inactivity) caused a significant response in RR and QT interval lengths, but not PR_seg (two-way RM ANOVA, Dunnet’s post hoc, difference from t = −5). E Acute change in RR (across 5-min analysis bins) was mirrored by concordant changes in QT, but not PR_seg. F Representative body temperature profile recorded across a 9-h advance of the LD cycle (data are double plotted; shaded regions indicate periods of darkness). G The 9-h phase advance led to profound separation in RR (black) and PR_seg (orange) interval rhythms. H Mean RR and PR_seg intervals measured across dark (green) and light (pink) phases the day prior to shift (day 0) and equivalent times on the day of the shift (day 1; two-way RM ANOVA, n = 11 mice). I Misalignment of RR and PR_seg rhythms in response to the shift in LD cycles disrupts the normal temporal relationship of the two parameters. Data reflect group mean PR_seg/RR on the day prior to shift (black) and the shift day (blue). All data presented as mean ± SEM. ns P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001. Source data and statistical details are provided as a Source Data File.

Fig. 4. Relative contribution of autonomic nervous system input and local cardiac clock in setting RR and PR_seg in mice.

A Heart rate variability (HRV) during autonomic blockade (with metoprolol and atropine) in conscious free-moving mice (left panel: a geometric measure of HRV, shaded region represents + /− SEM, right panels: Poincaré plots pre-blockade (baseline, bl) and under complete block). B Complete autonomic blockade (met + atr) reduced, but did not remove, time-of-day-dependent differences in RR or QT intervals (blue: ZT0; black: ZT12; n = 8 mice). In contrast, no time-of-day difference was observed in PR_seg under complete autonomic blockade (two-way RM ANOVA, Sidak’s post hoc). C In contrast to Bmal1^Fl/Fl control mice, no time-of- day differences in RR or QT intervals were evident in αMHC^CREBmal1^Fl/Fl mice following complete autonomic blockade (n = 7 Bmal^Fl/Fl, 9 aMHC^creBmal^Fl/Fl; two-way RM ANOVA, Sidak’s post hoc). D Twenty-four hours distribution of RR intervals in control (Bmal1^Fl/Fl; top) and littermate cardiomyocyte-specific Bmal1 knockout mice (αMHC^CREBmal1^Fl/Fl; bottom). Heatmap shows the occurrence of RR interval distribution across 5 days of recording; black line reflects group mean. Substantial RR interval lengthening and variability were evident in the day and late-night. All data presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 between time, ^#P < 0.05 between treatment. See Supplementary Fig. S8 for additional information related to this figure. Source data and statistical details are provided as a Source Data File.

Fig. 5. A local heart clock drives time-of-day-dependent rhythms in cardiomyocyte excitability.

A Representative recordings of spontaneous atrial (pink) and ventricular (green) electrograms recorded in ex vivo Langendorff-perfused hearts collected from control (Bmal1^Fl/Fl) and αMHC^CREBmal1^Fl/Fl mice at ZT0 or ZT12. B Interbeat interval showed a time-of-day (ZT0 blue, ZT12 black) difference in control but not αMHC^CREBmal1^Fl/Fl hearts; A–V conduction delay in isolated hearts did not vary by time of day (n = 7 hearts/group, two-way RM ANOVA, Sidak’s post hoc). C Representative multielectrode array recording of spontaneous action potential firing in primary cardiomyocytes. D, E Spontaneous firing rate in isolated cardiomyocytes showed robust circadian variation in constant culture conditions (D, n = 8, one-way RM ANOVA, Holm–Sidak post hoc), which followed rhythms of mPER2::LUC bioluminescence recorded in parallel cultures (RLU: relative light units, E). F Representative ventricular traces and pacing protocol (stimulation train shown below recording). Inset showing recovery to normal sinus rhythm (top) and induced ventricular tachycardia (VT; bottom). G Control hearts (n = 7/timepoint) showed a significant time-of-day susceptibility to VT (ZT0: 2 of 7 hearts tested, ZT12: 6 of 7; Chi-square), whereas only 1 of 14 αMHC^CREBmal1^Fl/Fl hearts (n = 7/timepoint) were susceptible. All data presented as mean ± SEM. *P < 0.05, **P < 0.01. Source data and statistical details are provided as a Source Data File.

Fig. 6. In vivo propensity for catecholamine-induced bidirectional VT displays diurnal rhythmicity.

A Typical ECG traces before (baseline) and after injection of caffeine and adrenaline. Animals displayed rapid tachycardia, typically followed by bradycardia and sinus pauses and/or premature complexes, with some progressing into bidirectional VT (bottom right). Vertical scale bars represent 0.5 mV and inset times are time from returning to the cage after injection. B Proportion of animals that display evidence of each stage of VT progression. At ZT12, all control animals (six of six mice tested) displayed robust evidence of bidirectional VT, compared to only one of six mice tested at ZT0. Only two of five αMHC^CREBmal1^Fl/Fl mice displayed bidirectional VT at ZT12. *P < 0.05, **P < 0.01, Chi-square. Source data and statistical details are provided as a Source Data File.