Inducibility, but not stability, of atrial fibrillation is increased by NOX2 overexpression in mice

Abstract Aims Gp91-containing NADPH oxidases (NOX2) are a significant source of myocardial superoxide production. An increase in NOX2 activity accompanies atrial fibrillation (AF) induction and electrical remodelling in animal models and predicts incident AF in humans; however, a direct causal role for NOX2 in AF has not been demonstrated. Accordingly, we investigated whether myocardial NOX2 overexpression in mice (NOX2-Tg) is sufficient to generate a favourable substrate for AF and further assessed the effects of atorvastatin, an inhibitor of NOX2, on atrial superoxide production and AF susceptibility. Methods and results NOX2-Tg mice showed a 2- to 2.5-fold higher atrial protein content of NOX2 compared with wild-type (WT) controls, which was associated with a significant (twofold) increase in NADPH-stimulated superoxide production (2-hydroxyethidium by HPLC) in left and right atrial tissue homogenates (P = 0.004 and P = 0.019, respectively). AF susceptibility assessed in vivo by transoesophageal atrial burst stimulation was modestly increased in NOX2-Tg compared with WT (probability of AF induction: 88% vs. 69%, respectively; P = 0.037), in the absence of significant alterations in AF duration, surface ECG parameters, and LV mass or function. Mechanistic studies did not support a role for NOX2 in promoting electrical or structural remodelling, as high-resolution optical mapping of atrial tissues showed no differences in action potential duration and conduction velocity between genotypes. In addition, we did not observe any genotype difference in markers of fibrosis and inflammation, including atrial collagen content and Col1a1, Il-1β, Il-6, and Mcp-1 mRNA. Similarly, NOX2 overexpression did not have consistent effects on RyR2 Ca2+ leak nor did it affect PKA or CaMKII-mediated RyR2 phosphorylation. Finally, treatment with atorvastatin significantly inhibited atrial superoxide production in NOX2-Tg but had no effect on AF induction in either genotype. Conclusion Together, these data indicate that while atrial NOX2 overexpression may contribute to atrial arrhythmogenesis, NOX2-derived superoxide production does not affect the electrical and structural properties of the atrial myocardium.


ECG and AF induction protocol
Surface ECG was obtained by connecting limb needle electrodes to an Iso-DAM8A amplifier (World Precision Instruments) and CED Power 1401-3A interface (Cambridge Electronic Design Ltd.). Data were acquired using the Spike 2 electrophysiology software. ECG signals were sampled at 2000 Hz and displayed in real time. After placement of the needle electrodes, mice were given a minimum of 5 minutes for the heart rate to stabilize. Standard ECG parameters were measured and averaged from five consecutive beats: R-R interval, P-Q interval, P wave duration, QRS duration, and corrected QT interval. The corrected QT interval (QTc) was calculated as per Bazett's formula (modified for the mouse): QTc = QT/√(RR/100).
AF was induced by atrial burst pacing via a transoesophageal octapolar catheter. First, the catheter was inserted into the oesophagus and placed at the location where the amplitude of the atrial signal was maximal. The diastolic pacing threshold was then determined by pacing the atria at a basic cycle length that was 10-to 20-ms under the sinus cycle length (typically 100ms) while slowly changing the position of the catheter and the amperage to find the position with the lowest atrial capture threshold (i.e., the lowest voltage required to achieve 1:1 conduction to the ventricles), which also indicates correct catheter placement at the atria. All subsequent stimulations were delivered at twice the threshold amperage with a stimulus duration of 1-ms. The occurrence of AF was identified from the surface ECG by the development of rapid and irregular atrial rhythms, often characterised by the absence of regular P waves, and irregular ventricular activation. AF was then analysed based on arrhythmia duration, initially applying a cut-off of 2-sec and subsequently 5-sec and 10-sec. If burst electrical stimulation provoked arrhythmias that were less than the specified cut-offs or did not evoke an arrhythmic episode at all, the mouse was said to be in sinus rhythm. AF vulnerability was assessed by quantifying the incidence (proportion of mice that developed an episode of NOX2 activity is not causal for atrial fibrillation 5 AF) and probability (number of arrhythmic episodes divided by the total number of testing manoeuvres applied) of pacing-induced AF. AF duration was measured from the end of the pacing train until the first sinus P wave and maximum (the longest AF episode in each animal), cumulative (cumulative sum of all discrete AF episodes in each animal), and mean (average of all discrete AF episodes in each animal) AF durations were quantified and compared between groups.

Data collection and offline analysis
Image sequences were processed using the GView software. The graphical interface allows users to select regions of interest on the image and generates a time series of optical action potentials that is exported as plain text. Optical action potentials were obtained from five different regions (12-by-12 pixels) and APD values were computed from inverted fluorescence data using custom-written MATLAB scripts (version R2015a, Mathworks) that were developed based on the methods and recommendations by Laughner et al. 35 Briefly, baseline drift in the raw optical recordings was corrected by fitting a 5 th -order polynomial and subtracting that from the raw signal to establish a constant baseline level. A bi-directional filter (5 th order Butterworth filter) with a 100 Hz cut-off was then applied to the fluorescence trace to reduce high frequency noise (>100 Hz) and increase the signal-to-noise ratio. Next, individual action potentials were identified using the findpeaks function which locates the local maxima within the trace. To ensure the software did not detect unwanted peaks buried in noise, a minimum peak interval and amplitude was defined. Action potentials were then analysed to determine the amplitude, and activation and repolarization times. The amplitude was given by the difference between the maximum value and the baseline of the fluorescence trace which was an average of 50 frames preceding the action potential upstroke. The activation time was determined as the time of the first derivative of the fluorescence signal (dF/dt) corresponding to the steepest segment, or the maximum slope, of the action potential upstroke. Repolarization time was measured as a time when a specified level of repolarization (30%, 50% and 80% [APD30, APD50, and APD80, respectively]) was reached during the repolarization phase of the action potential. APD was calculated as the difference between repolarization and activation times and averaged to give an estimate of the global APD in each atrium.
Epicardial conduction velocities were analysed from optical image sequences using the Ccoffinn toolkit implemented in MatLab (courtesy of Jakub Tomek, University of Oxford). 36 Ccoffinn is a tool that constructs a representation of cardiac waves and tracks their movement, using that information to describe the properties of the observed waves. Briefly, on a frame by frame basis, a peak-finding algorithm identifies the location of pixels that are activated in each frame of the recording (to estimate which parts of the tissue are firing in a synchronized manner) and groups them together as wavefronts. From this, waves are detected frame by frame and the movement of wavefronts is tracked throughout the recording. The tracking information (speed, distance, and direction) is then used to extract the conduction velocity (CV) of each wavefront. The mean CV, taken as the average of all wavefronts in the activation sequence, and maximum CV, taken as the fastest wavefront in the activation sequence, were calculated and averaged for each atrium during sinus rhythm.

Isolation of atrial myocytes
Mice were injected with heparin (1000 U/mL, ip) and sacrificed by Schedule 1 cervical dislocation. The thorax was opened by a midsternal incision and the heart was dissected out and rapidly immersed in a Ca 2+ -free Tyrode solution (in mmol/L: 130 NaCl, 5.6 KCl, 3.5 MgCl2, 5.0 HEPES, 0.4 Na2PO4, 20 taurine, 10 glucose, with pH adjusted to 7.4 with NaOH).
NOX2 activity is not causal for atrial fibrillation 7 The lungs, thymus, and fat were carefully removed to expose the aorta which was subsequently inserted onto a cannula (made from a blunted 23 gauge needle) filled with the Ca 2+ -free solution. A surgical suture thread (Ethicon 5/0 silk) was looped around the aorta and tightened in order to secure the cannula. A second suture was added to ensure adequate filling and perfusion of the atria and the aorta was then cannulated onto a Langendorff perfusion apparatus.
The heart was perfused with a Ca 2+ -free Tyrode solution (maintained at 37 °C) for 3 minutes to clear the blood, and then with a primary enzyme solution (containing 1 mg/mL collagenase type II, 0.233 mg/mL protease, 0.166% BSA, and 50 µmol/L Ca 2+ in Tyrode solution) for a further 8-10 minutes until the heart was digested. At the end of this period, the heart was  Likewise, the RyR2 phosphorylated fraction at Ser2814 did not differ between genotypes (for the RA, n = 17 for both groups, P >0.05 and for the LA n=16 for WT and n=18 for NOX2-Tg, P >0.05). Top panels include representative Western blots from each atrium and genotype. Graphs show individual data points with medians and IQRs.