Pro‐arrhythmic atrial phenotypes in incrementally paced murine Pgc1β −/− hearts: effects of age

New Findings What is the central question of this study? Can we experimentally replicate atrial pro‐arrhythmic phenotypes associated with important chronic clinical conditions, including physical inactivity, obesity, diabetes mellitus and metabolic syndrome, compromising mitochondrial function, and clarify their electrophysiological basis? What is the main finding and its importance? Electrocardiographic and intracellular cardiomyocyte recording at progressively incremented pacing rates demonstrated age‐dependent atrial arrhythmic phenotypes in Langendorff‐perfused murine Pgc1β −/− hearts for the first time. We attributed these to compromised action potential conduction and excitation wavefronts, whilst excluding alterations in recovery properties or temporal electrophysiological instabilities, clarifying these pro‐arrhythmic changes in chronic metabolic disease. Atrial arrhythmias, most commonly manifesting as atrial fibrillation, represent a major clinical problem. The incidence of atrial fibrillation increases with both age and conditions associated with energetic dysfunction. Atrial arrhythmic phenotypes were compared in young (12–16 week) and aged (>52 week) wild‐type (WT) and peroxisome proliferative activated receptor, gamma, coactivator 1 beta (Ppargc1b)‐deficient (Pgc1β −/−) Langendorff‐perfused hearts, previously used to model mitochondrial energetic disorder. Electrophysiological explorations were performed using simultaneous whole‐heart ECG and intracellular atrial action potential (AP) recordings. Two stimulation protocols were used: an S1S2 protocol, which imposed extrasystolic stimuli at successively decremented intervals following regular pulse trains; and a regular pacing protocol at successively incremented frequencies. Aged Pgc1β −/− hearts showed greater atrial arrhythmogenicity, presenting as atrial tachycardia and ectopic activity. Maximal rates of AP depolarization (dV/dt max) were reduced in Pgc1β −/− hearts. Action potential latencies were increased by the Pgc1β −/− genotype, with an added interactive effect of age. In contrast, AP durations to 90% recovery (APD90) were shorter in Pgc1β −/− hearts despite similar atrial effective recovery periods amongst the different groups. These findings accompanied paradoxical decreases in the incidence and duration of alternans in the aged and Pgc1β −/− hearts. Limiting slopes of restitution curves of APD90 against diastolic interval were correspondingly reduced interactively by Pgc1β −/− genotype and age. In contrast, reduced AP wavelengths were associated with Pgc1β −/− genotype, both independently and interacting with age, through the basic cycle lengths explored, with the aged Pgc1β −/− hearts showing the shortest wavelengths. These findings thus implicate AP wavelength in possible mechanisms for the atrial arrhythmic changes reported here.


Introduction
Arrhythmogenesis is a complex physiological phenomenon of dysregulated cardiac electrical activity, with both short-and long-term consequences. The complex embryological origins and anatomical structure of the atria make them particularly vulnerable to arrhythmic syndromes. Atrial fibrillation (AF) is of particular clinical importance, affecting 1-3% of individuals in Western countries (Friberg & Bergfeldt, 2013). The processes underpinning its induction and maintenance remain incompletely explained, but involve complex interactions between altered cardiac electrical properties and functional changes in the atria, which occur over the short and long term. It has been suggested that AF is a self-perpetuating process, triggered initially by focal ectopic activity arising in the pulmonary veins that drive cumulative electrical and structural remodelling processes, themselves generating arrhythmic substrate (Haïssaguerre et al. 1998).
These changes are exacerbated by several interacting upstream factors, with ageing and metabolic disease central to a number of these. There is a pronounced increase in the prevalence of AF with age, from ß4% of individuals aged 60-70 years to nearly 20% of individuals ࣙ80 years (Zoni-Berisso et al. 2014). Likewise, metabolic factors may explain ß60% of current upward trends in incidences of AF (Miyasaka et al. 2006). Metabolic disease and obesity have been implicated as risk factors, themselves age dependent, for ventricular arrhythmias (Adabag et al. 2015). Likewise, the risk of AF increases with physical inactivity (Mozaffarian et al. 2008), obesity (Tedrow et al. 2010), diabetes mellitus (Nichols et al. 2009) and metabolic syndrome (Watanabe et al. 2008). Amelioration of metabolic disease improves both risk profiles and responses to therapy (Tedrow et al. 2010). Furthermore, it has been shown that manipulation of key components of cellular energy production pathways suppresses arrhythmia in known arrhythmogenic models (Liu et al. 2009).
Mitochondrial function may be integral to relationships between ageing, metabolism and arrhythmia. Mitochondria provide >95% of the ATP required for Ca 2+ homeostasis and maintenance of transmembrane ionic gradients in addition to cardiac muscle contraction (Barth & Tomaselli, 2009). Abnormal mitochondrial structure and function have been reported in animal models of AF (Ausma et al. 1997). Analysis of mitochondria from cardiomyocytes of human AF patients demonstrates increased DNA damage (Tsuboi et al. 2001;Lin et al. 2003), structural abnormalities (Bukowska et al. 2008) and impaired function . Furthermore, a number of targeted mitochondrial DNA mutations accumulate with age and show increased incidences in AF (Lai et al. 2003). Mitochondrial dysfunction also results in generation of reactive oxygen species, which has been implicated in the pathogenesis of human AF (Korantzopoulos et al. 2007). Experiments producing acute mitochondrial impairment through ischaemia-reperfusion correspondingly demonstrate pro-arrhythmic regional heterogeneities in ventricular action potential duration (APD) and arrhythmogenesis suppressed by pharmacological manipulations of the mitochondrial membrane potential (Akar et al. 2005; see also Brown et al. 2010). These pro-arrhythmic changes were successfully suppressed by pharmacological manipulations of the mitochondrial membrane potential (Akar et al. 2005). Further work demonstrated that stabilization of the mitochondrial membrane potential also inhibits arrhythmias in other settings (Brown et al. 2010).
However, few experiments have explored the effects of chronic age-dependent energetic deficiency arising from ageing and mitochondrial dysfunction on the generation of atrial arrhythmias or determined their underlying electrophysiological abnormalities. The present study used a peroxisome proliferative activated receptor, gamma, coactivator 1 beta (Ppargc1b; hereafter Pgc1β)-deficient (Pgc1β −/− ) model (Mouse Genome Informatics identification: MGI:2444934 (http://www. informatics.jax.org/)) that has been used previously to study the biochemical consequences of metabolic energetic deficiency. This family of transcriptional coactivators provides central regulators of cellular and mitochondrial function controlling key metabolic pathways, including fatty acid oxidation, mitochondrial biogenesis and oxidative phosphorylation. It occurs abundantly in energetically active tissue, such as the heart . Reduced Pgc1 levels and impaired mitochondrial function occur in diabetes mellitus, the metabolic syndrome, ageing and heart failure (Garnier et al. 2003;Leone & Kelly, 2011). Despite normal baseline function, Pgc1β-deficient mice demonstrate compromised heart rate responses with adrenergic stimulation (Lelliott et al. 2006). Their ex vivo Langendorff-perfused hearts show ventricular arrhythmic tendencies, abnormal diastolic Ca 2+ transients and altered ion channel expression (Gurung et al. 2011). However, neither their atrial arrhythmic phenotypes nor their accompanying electrophysiological changes have been explored.
The present study complements previous reports in murine hearts carrying genetic abnormalities in specific ion channels modelling ventricular arrhythmic conditions (Huang, 2017). The arrhythmic substrate in these different exemplars was variously identified with altered action potential (AP) initiation and conduction (Martin et al. 2011;Ning et al. 2016), AP recovery (Sabir et al. 2007b) and arrhythmic triggers (Thomas et al. 2007;Goddard et al. 2008). The present experiments, likewise, looked for the presence of arrhythmic phenotypes provoked by the imposition of extrasystolic S2 stimuli at differing S1S2 intervals following trains of regular S1 pacing as well as steady-state pacing at progressively decreased basic cycle lengths (BCLs). The findings were then matched to results from simultaneous determinations of AP activation and recovery, as well as temporal instabilities in the form of AP alternans (Matthews et al. 2010), APD-diastolic interval (DI) restitution relationships (Kim et al. 2002;Matthews et al. 2012) and spatiotemporal indicators of AP wavelength. This study represents the first to make such measurements in murine atria. Studies in human AF have reported increased slopes in restitution plots of APD-DI, intervening between full AP recovery and the peak of the subsequent AP. However, these studies did not look for accompanying alternans (Kim et al. 2002). However, the present findings do support previous reports associating AF with paradoxical suppression of APD alternans in a vagally mediated canine AF model (Lu et al. 2011).

Animals and ethical approval
This research was regulated under the Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012 following ethical review by the University of Cambridge Animal Welfare and Ethical Review Body (Home Office PPL no. 70/8726), and followed recommendations provided by the Physiological Society (Grundy, 2015).
Mice were housed in an animal facility with 12 h-12 h light-dark cycles at a temperature maintained at 21°C. Animals were fed sterile chow (RM3 Maintenance Diet; SDS, Witham, UK) and had free access to water, bedding and environmental stimuli. The C57/B6 mice strain was used as the background for both wild-type (WT) and Pgc1β −/− mice, generated using a triple LoxP targeting vector as previously described (Lelliott et al. 2006). Four experimental groups were studied: young WT (n = 20), young Pgc1β −/− (n = 23), aged WT (n = 22) and aged Pgc1β −/− (n = 22). All young mice were aged between 12 and 16 weeks and aged mice >52 weeks. Mice were administered 200 IU of unfractionated heparin (Sigma-Aldrich, Poole, UK) in the intraperitoneal space before being killed by cervical dislocation [Schedule 1; Animals (Scientific Procedures) Act 1986]. No recovery, anaesthetic or surgical procedures were required.

Whole-heart intracellular microelectrode recordings
Animals were killed for rapid sternectomy and cardiectomy. Hearts were cannulated and secured as previously described (Zhang et al. 2011;Matthews et al. 2012;Ning et al. 2016). No obvious macroscopic defects were observed in any heart. Hearts were then mounted on a horizontal Langendorff apparatus that was electrically insulated and incorporated into an intracellular rig within a Faraday cage, incorporating a light microscope (objective ×5, eyepiece ×5; W. Watson and Sons Limited, London, UK), organ bath, custom-built microelectrode amplifier and head stage.
To facilitate impalement of the left atrium, hearts were mounted in a standard anatomical position to allow pacing from the right atrium (RA) and access to the left atrium (LA). The left atrium was displaced posteriorly and held in place with three A1 insect pins. The positions of the recording and stimulating electrodes were controlled by two precision micromanipulators (Prior Scientific Instruments, Cambridge, UK). In all experiments, the stimulating electrode was consistently positioned at the posterior aspect of the RA, and recordings were made from the central region of the LA, minimizing variability in distances between the respective electrodes, so that alterations in AP latencies therefore provided indications of corresponding conduction velocity changes. Hearts were perfused with plain KH solution (flow rate of 2.05 ml min −1 ) to allow establishment of a regular intrinsic rhythm. Preparations with an intrinsic rate of <5 Hz or which did not display 1:1 atrioventricular conduction for >10 min postperfusion were not used for experimentation. Perfusion with KH containing 20 μM blebbistatin was then commenced until motion was adequately minimized before resumption of perfusion with plain KH solution.
A microelectrode (tip resistance 15-25 M ) was pulled by a custom-built microelectrode puller from a glass pipette (1.2 mm o.d., 0.69 mm i.d.; Harvard Apparatus, Cambourne, UK), filled with 3 M KCl and then inserted into a right-angled microelectrode holder (Harvard Apparatus). The recorded signal from the microelectrode was passed through a head stage for preamplification forming part of a highinput impedance direct-current microelectrode amplifier system (University of Cambridge, Cambridge, UK) before bandpass filtering (between 0 and 2 kHz) and analog-to-digital conversion at a sampling frequency of 10 kHz (1401; Cambridge Electronic Design, Cambridge, UK). A successful impalement was identified by the abrupt appearance of a resting membrane potential more negative than −70 mV and regular APs with a stable waveform. A Ag/AgCl electrode in contact with the bath solution was used as the reference electrode for all measurements.

Whole-heart ECG recordings
Correlation between electrical signals at the level of the whole organ and intracellular voltage recordings was permitted by simultaneous ECG recordings from the Langendorff-perfused hearts. Two unipolar ECG electrodes were placed at fixed positions adjacent to the heart in the organ bath that corresponded to the standard three-lead ECG co-ordinates. The recorded signals were passed through head stages for preamplification before amplification (Neurolog NL104 amplifier), bandpass filtering (between 5 and 500 Hz; NL 125/6 filter; Digitimer, Welwyn Garden City, UK) and digitization at a sampling frequency of 10 kHz (1401; Cambridge Electronic Design).

Pacing protocols
Hearts were stimulated at an amplitude of twice diastolic threshold voltage plus 0.5 mV. Hearts underwent two separate pacing protocols during each experiment. First, a standardized S1S2 protocol was used to determine the atrial effective refractory period (AERP) from the ECG recordings. This delivered successive trains of eight S1 stimuli separated by an interval of 125 ms, followed by a solitary extrasystolic stimulus (S2) initially delivered 89 ms after the preceding S1 stimulus. This pattern of stimulation was repeated with the S1S2 interval decremented by 1 ms for each successive cycle until failure of stimulus capture. Incremental pacing protocols then began after achieving stable microelectrode impalement. These consisted of cycles of regular pacing, each of 100 stimulations. They began with a BCL of 130 ms that was then decremented by 5 ms for each subsequent cycle. These were repeated until the heart entered into 2:1 block or sustained arrhythmia.

Data and statistical analysis
Data captured by the Spike2 software package (Cambridge Electronic Design) was analysed using a custom-written program using the python programming language. Alternans was defined as an occurrence of alternating beat-to-beat changes in the value of a parameter such that the direction of the change oscillates for at least 12 successive action potentials. Statistical analysis was carried out using the R programming language (R Core Team, 2015) and plots with the grammar of graphics package (Wickham, 2009). All data are expressed as means ± SD, and a P value of < 0.05 was taken to be significant. Different experimental groups were compared with a two-factor ANOVA. F values that were significant for interactive effects prompted post hoc testing with Tukey's honest significant difference testing. If single comparisons were made, Student's two-tailed t test was used to compare significance. Categorical variables were compared using Fisher's exact test. Kaplan-Meier estimates were compared with the log rank test.

Aged Pgc1β −/− hearts develop a pro-arrhythmic phenotype
Electrocardiographic recordings were first made through the S1S2 protocol. Extrasystolic (S2) stimuli were interposed at successively shorter intervals following trains of eight regular (S1) stimuli applied at a 125 ms basic cycle length. This explored for the presence and the frequency of arrhythmic phenotypes in the intact ex vivo Langendorff-perfused hearts. Figure 1 shows typical ECG recordings from aged Pgc1β −/− hearts at a slow time base during an S1S2 stimulation protocol. These include episodes of premature atrial complexes following successive S2 stimuli and a short run of atrial tachycardia captured during a typical stimulus train at the end of an S1S2 protocol (Fig. 1A). On an expanded time base, these could be characterized by spontaneous atrial P waves (dashed arrow) in contrast to the paced P waves (continuous downward-pointing arrows) following imposed pacing spikes (Fig. 1B, arrowed) between successive ventricular complexes (upward-pointing arrows). Some protocols also elicited runs of atrial tachycardia (Fig. 1C). The S1S2 interval at the onset of failure of stimulus capture made it possible to determine the AERP corresponding to the specific (8 Hz) pacing rate. A comparison of AERPs obtained from the S1S2 protocol in young WT (24.8 ± 5.8 ms), aged WT (28.8 ± 6.1 ms), young Pgc1β −/− (29.3 ± 3.8 ms) and aged Pgc1β −/− hearts (30.1 ± 8.9 ms) demonstrated no significant differences between groups and provided indications of the extent to which BCLs could be decreased in the succeeding incremental pacing experiments. We then performed simultaneous whole-heart ECG recordings and intracellular AP measurements from single cardiomyocytes after achieving stable microelectrode impalements, with consistent stimulating and recording electrode positions. The intracellular AP recordings provided accurate measurements of AP characteristics related to AP initiation, activation and recovery. These included maximal AP upstroke rates (dV/dt max ), AP latencies, defined as the time interval between the pacing spike of each AP and the maximal deflection, or peak, of the AP, AP durations at 90% recovery (APD 90 ), and resting membrane potentials (RMPs). The measurements of dV/dt max and RMPs would not have been available with the monophasic action potential electrode methods used on previous occasions (Sabir et al. 2007a(Sabir et al. , 2008a. The incremental pacing protocols applied cycles of 100 regular pacing stimuli at successively decremented BCLs. Figure 2 illustrates results of the subsequent incremental pacing procedure comparing ECG (i) and intracellular traces (ii) in conditions of regular activity ( Fig. 2A) and occurrences of premature atrial complexes (Fig. 2B), atrial fibrillation ( Fig. 2C) and APD 90 alternans ( Fig. 2D) in an aged Pgc1β −/− heart. Individual hearts subjected to incremental pacing could therefore display phenotypes that differed from the results of extrasystolic (S2) stimuli.
The cycles of incremental pacing continued with decreasing BCLs until the onset of either 2:1 capture or arrhythmia was reached. Kaplan-Meier curves plotting the probability of 1:1 capture of the groups as a function of BCL (Fig. 3) demonstrated a progressive reduction in the number of hearts continuing to show 1:1 capture at BCLs shorter than ß70 ms. This would reflect their refractory properties at the steady-state pacing frequencies close to this cut-off. The statistical analysis to follow will therefore analyse data for parameters at BCLs no shorter than ß50 ms. A log rank test confirmed that the survival curves were in fact significantly different (P = 0.0028). Young WT hearts showed fall-offs at shorter BCLs than in the remaining groups and thus could be paced at higher frequencies than the other hearts, including aged WT hearts. Table 1 summarizes the incidences of arrhythmic phenomena, whether in the form of atrial tachycardia or ectopic atrial deflections, through both pulse protocols. These together suggested a more marked pro-arrhythmic phenotype in aged Pgc1β −/− hearts than in the remaining groups. With the S1S2 pulse protocol, the incidence of ectopic deflections was similar between groups, but the incidence of atrial tachycardias was greater in aged Pgc1β −/− hearts than in young WT, aged WT or young Pgc1β −/− hearts, which all showed similar incidences of atrial tachycardias (P = 0.043). Hearts displaying arrhythmias with extrasystolic (S2) provocation often failed to show arrhythmias on incremental pacing, and a relatively small number (three) of hearts

. Electrocardiographic and intracellular recordings during incremental pacing
A, typical recordings from incremental pacing protocols illustrating steady-state ECG recordings (i) from a wild-type (WT) heart as well as the corresponding intracellular action potential (AP) recordings (ii). Ectopic activity (arrow) from an showed arrhythmia amongst the experimental groups. The incremental pacing protocol thus resulted in few incidences of atrial tachycardia. Nevertheless, incidences of ectopic deflections were then greater in aged Pgc1β −/− hearts than in young WT, aged WT and young Pgc1β −/− hearts (P = 0.041).

Altered atrial AP characteristics in young and aged
Pgc1β −/− hearts Figure 4 summarizes the activation and recovery characteristics of APs obtained through the incremental pacing procedures. It illustrates the corresponding alterations in dV/dt max (Fig. 4A), AP latencies (Fig. 4B), APD 90 (Fig. 4C), RMP (Fig. 4D) and diastolic intervals intervening between the attainment of 90% of AP recovery and the peak of the subsequent action potential (DI 90 ; Fig. 4E) with alterations in BCL in young and aged Pgc1β −/− and WT hearts. These parameters varied in a approximately linear manner with BCL, and their overall magnitudes could be compared by the areas beneath their curves. These are summarized in Table 2 which, in addition to showing results from the individual experimental groups of young and aged WT and Pgc1β −/− hearts, also provides the results of grouping all aged, young as well as WT and Pgc1β −/− atria for the statistical analysis. Two-way ANOVA demonstrated that the Pgc1β −/− hearts displayed decreased dV/dt max (P = 0.000020) and lower APD 90 values (P = 0.00018) compared with WT hearts; there were no variations in RMP between groups. Genotype and age exerted significant interacting effects on DI 90 and AP latency values (P = 0.0081, P = 0.043, respectively). Post hoc tests did not reveal further statistically significant differences.
Reduced temporal heterogeneities in atrial AP characteristics in aged Pgc1β −/− hearts Instabilities in characteristics of successive APs, often taking the form of episodes of alternans, presage major ventricular arrhythmias in clinical situations. They have been described in experimental conditions as alternating variations in temporal properties of AP excitation and/or recovery that occur with varying heart rates in analyses of pro-arrhythmic tendencies associated with ventricular arrhythmogenesis (Sabir et al. 2007a(Sabir et al. , 2008a. We sought to analyse whether such phenomena are important in atrial arrhythmogenesis. Figure 5 summarizes the incidences of such alternans in the activation parameters dV/dt max aged Pgc1β −/− heart is shown in B, with the corresponding ECG (i) and intracellular AP recordings (ii). C, evidence of atrial fibrillation in the ECG (i) recording of an aged Pgc1β −/− heart, with concurrent intracellular AP recording (ii). D illustrates AP duration alternans in an AP recording from a young WT heart. ( Fig. 5A) and AP latency (Fig. 5B) and the recovery parameters APD 90 (Fig. 5C) and RMP (Fig. 5D) in young and aged WT and Pgc1β −/− hearts through the incremental pacing protocol. Overall incidences of alternans were assessed by summing the individual incidence of alternans at each BCL. The different groups showed similar distributions in the occurrence of alternans at different BCLs. However, statistical comparisons of the overall incidences of alternans throughout the entire range of BCLs indicated that Pgc1β −/− hearts have reduced incidences and durations of alternans episodes. Thus, two-way ANOVA demonstrated that there were no significant effects of genotype on the incidence of alternans in dV/dt max , APD 90 , AP latency or RMP. Ageing independently reduced the incidence of alternans in aged compared with young hearts (dV/dt max 24 ± 19.9 versus 39 ± 19.7 beats, P = 0.0013; APD 90 , 37 ± 19.9 versus 50 ± 19.7 beats, P = 0.0051; AP latency, 53 ± 19.9 versus 68 ± 19.7 beats, P = 0.00045; and RMP, 30 ± 13.2 versus 40 ± 13.1 beats; P = 0.0011). Age and genotype exerted interacting effects on the incidence of AP latency alternans (P = 0.032). Post hoc testing demonstrated less AP latency alternans in aged than young Pgc1β −/− hearts (47 ± 23.5 versus 71 ± 14.4 beats, P = 0.00041) and in aged Pgc1β −/− than young WT hearts (47 ± 23.5 versus 64 ± 22.4 beats, P = 0.036).
Likewise, two-way ANOVA demonstrated that the magnitude of dV/dt max , APD 90 and RMP (although not AP latency) alternans, reflected in the areas under the respective curves, was smaller in aged than young hearts (P = 0.037, P = 0.038, P = 0.052 and P = 0.066, respectively; Fig. 6). There were no effects of genotype or interacting effects of genotype and age together on the overall magnitudes of oscillation. The total numbers of episodes of APD 90 , AP latency or RMP alternans were indistinguishable between groups, whilst Pgc1β −/− hearts showed fewer episodes of dV/dt max alternans than WT hearts (8.54 ± 5.0 versus 13.37 ± 9.5 episodes of alternans, P = 0.0021) and aged hearts fewer episodes of dV/dt max alternans than young hearts (9.24 ± 7.5 versus 12 ± 7.5 episodes of alternans, P = 0.030).

Spatiotemporal representations of AP excitation in
In contrast, analysis of spatial representations of action potential activation, as opposed to the temporal characteristics of AP recovery considered above, replicated the arrhythmic phenotypes in terms of the accompanying electrophysiological abnormalities particularly in aged Pgc1β −/− hearts. Here, the AP travelling waves were represented in terms of wavelengths of excited tissue undergoing APs (λ) and resting wavelengths (λ 0 ) of the succeeding tissues that had completed their subsequent AP recovery. These λ and λ 0 terms were obtained by multiplying 1/(AP latency), reflecting conduction velocity, by the corresponding APD 90 or DI 90 values, respectively (Matthews et al. 2013b). These were then compared in young and aged WT and Pgc1β −/− atria through the different BCLs examined. Areas under plots of λ against BCL (Fig. 7B) then demonstrated that the Pgc1β −/− as opposed to the WT genotype, but not age, independently (two-way ANOVA: P = 0.6 × 10 −4 and 0.14, respectively) reduced λ. Additional, interacting effects (P = 0.048) through the range of explored BCLs, were reflected in the shorter λ values in both young (Tukey's test: P = 0.0001) and aged Pgc1β −/− (Tukey's test: P = 0.0008) compared with young WT atria. Likewise, λ values from the experimental groups all declined and converged with declining λ 0 and with shortening BCL, as previously reported (Matthews et al. 2013a;Ning et al. 2016b; Fig. 7C). Nevertheless, λ values in aged Pgc1β −/− hearts consistently fell below those in remaining groups (n = 14 points, sign test, P < 0.01). Accordingly, areas beneath the curves reflected significantly greater λ at the longer (85-130 ms; two-way ANOVA: P = 0.042) but not the shorter (<85 ms) BCLs, resulting from interacting effects of age and Pgc1β −/− genotype (Matthews et al. 2013a;Ning et al. 2016b).

Discussion
Both age and energetic dysfunction are known risk factors for atrial fibrillation (Go et al. 2001;Menezes et al. 2013).

H. Valli and others
The experimental basis for this effect was studied in young and aged Pgc1-deficient hearts used in previous biochemical studies of mitochondrial dysfunction. Pgc1 upregulates mitochondrial function and cellular energy homeostasis Finck & Kelly, 2006), acting on genes involved in fatty acid oxidation and electron transport (Arany et al. 2005). Both ageing WT (Froehlich et al. 1978;Lakatta & Sollott, 2002;Hatch et al. 2011;Yang et al. 2015) and energetically deficient Pgc1β −/− hearts show abnormalities in a range of cellular mechanisms with potential electrophysiological consequences for arrhythmia. Their increased production of reactive oxygen species (Grivennikova et al. 2010) modifies maximal voltage-dependent Na + and K + current (Wang et al. 2004;Liu et al. 2010), sarcolemmal K ATP channel function, Na + and Ca 2+ channel inactivation, and late Na + current and ryanodine receptor function (Huang, 2017). The associated ATP/ADP depletion opens sarcolemmal ATP-sensitive K + channels (sarcK ATP ; Akar & O'Rourke, 2011). Additional abnormalities in Ca 2+ homeostasis and delayed after-depolarization events potentially initiate pro-arrhythmic triggering activity (Gurung et al. 2011). These mechanisms could in turn potentially alter cell-cell coupling (Smyth et al. 2010), AP conduction (Liu et al. 2010), repolarization and refractoriness (Wang et al. 2004), as well as predispose to alternans and Ca 2+ -mediated pro-arrhythmic triggering phenomena (Terentyev et al. 2008).
Recent studies have reported both compromised heart rate responses to adrenergic stimulation (Lelliott et al. 2006) and pro-arrhythmic ventricular phenotypes in intact perfused murine Pgc1β −/− hearts (Gurung et al. 2011). However, the corresponding atrial electrophysiological AP or arrhythmic phenotypes have not been investigated. In the present study, we applied cellular electrophysiological recordings in cardiomyocytes within intact, normally functioning Langendorff-perfused hearts as opposed to isolated cardiomyocytes. We thereby determined steady-state arrhythmic and related restitution properties of murine atrial tissue for the first time.
Our experimental approach permitted simultaneous study of atrial arrhythmic properties of the whole heart and electrophysiological properties of single atrial cells. Aged Pgc1β −/− hearts demonstrated a significantly greater incidence of arrhythmogenic phenotypes compared with all the remaining groups. The stimulation procedures that applied extrasystolic S2 stimuli resulted in a higher incidence of atrial tachycardias, whereas those applying incremental increases in steady heart rates resulted in a higher incidence of ectopic atrial events. The latter suggests that atrial cardiomyocytes have a greater capacity for rapid pacing without producing pro-arrhythmic phenomena.
The association of the arrhythmic phenotype with alterations in the corresponding AP characteristics was then examined. Of the measured AP parameters, resting membrane potentials remained uniform throughout all experimental groups, consistent with clinical findings in atrial fibrillation (Bosch et al. 1999). The statistically most noticeable alterations involved compromised AP activation, which has been implicated in arrhythmic substrate on previous occasions King et al. 2013a;Huang, 2017). Thus, Pgc1β −/− atria displayed decreased dV/dt max compared with WT atria, in the absence of statistical effects of age or interactions between age and genotype. The dV/dt max correlates with peak Na + current, which in turn markedly influences AP conduction velocity to an extent dependent upon the conductivity between cells (Hunter et al. 1975;Hondeghem & Katzung, 1977;Usher-Smith et al. 2006;Fraser et al. 2011). Although young and aged Pgc1β −/− atria did not demonstrate significant differences in dV/dt max , AP latency measurements reflecting conduction velocity were influenced by interactions between age and genotype. They therefore account for the more marked pro-arrhythmic phenotype in the aged than the young Pgc1β −/− atria.
These findings are compatible with hypotheses relating reduced dV/dt max and increased AP latency to decreased peak Na + current in the conditions of agedependent mitochondrial dysfunction in Pgc1β −/− hearts (Hondeghem & Katzung, 1977;Usher-Smith et al. 2006;Fraser et al. 2011). Established reports associate metabolic insufficiency with reduced Na + channel function in a number of circumstances. First, in addition to reduced provision of ATP, disrupted mitochondrial activity increases production of reactive oxygen species (Manning et al. 1984;Fosset et al. 1988;Faivre & Findlay, 1990) and perturbs cytosolic NAD + /NADH. Both factors are implicated in altered Na + channel function in metabolically stressed cardiomyocytes (Liu et al. 2009) and are rescued by the mitochondrial reactive oxygen species scavenger mitoTEMPO (Liu et al. 2010).
Comparison of the present findings of reduced dV/dt max and increased AP latency in cardiomyocytes in intact tissue with the normal or even enhanced Na + currents in Pgc1β −/− cardiomyocytes subject to patch-clamping involving Ca 2+ chelation by intrapipette BAPTA (Gurung et al. 2011) are compatible with a mechanism of action involving the latter, acute effects, of altered Ca 2+ homeostasis upon membrane excitability. This could involve Ca 2+ -Na v 1.5 interactions involving direct Ca 2+ -Na V 1.5 binding at an EF hand motif close to the Na V 1.5 C-terminal (Wingo et al. 2004) or indirect Ca 2+ binding involving additional 'IQ' domain binding sites for Ca 2+ -calmodulin in the Na V 1.5 C-terminal region (Mori et al. 2000;Wagner et al. 2011;Grandi & Herren, 2014). If so, these findings in a model for metabolic disturbance provide a further example of potentially important effects of intracellular Ca 2+ homeostasis on arrhythmic substrate through altering AP propagation as a result of acute effects upon Na + channel function.
The electrophysiological alterations of atrial cardiomyocytes presented in this study led to investigations to distinguish mechanisms for the resulting arrhythmic substrate at the tissue level (Martin et al. 2012), relating these to previous reports from monogenic murine arrhythmic models. First, arrhythmic syndromes primarily attributed to repolarization abnormalities have been exemplified by murine Scn5a+/KPQ hearts. Arrhythmic substrate was there associated with altered relationships between APD 90 and effective refractory period with varying diastolic interval (DI 90 ; Nolasco & Dahlen, 1968;Sabir et al. 2008a;Matthews et al. 2010Matthews et al. , 2012 increasing the steepness of restitution curve plots of APD 90 against DI 90 at short BCLs. The consequent APD 90 instabilities would increase frequencies and amplitudes of alternans, culminating in arrhythmic substrate (Huang, 2017).
However, the present findings did not implicate such alternans in the arrhythmic instability in aged Pgc1β −/− atria. Atrial alternans was observed in AP trains during incremental steady-state pacing. However, its incidence was not increased by age or Pgc1β −/− genotype. Indeed, aged atria showed decreased incidences of alternans compared with young hearts. The Pgc1β −/− and aged atria showed fewer episodes of alternans than WT and young atria, respectively. Furthermore, it was the restitution curves of aged Pgc1β −/− hearts that showed the most reduced limiting slopes. These findings parallel previous reports in the vagally induced model of canine AF that likewise showed paradoxically less alternans and a flatter restitution slope than in the non-arrhythmic control state (Lu et al. 2011). Atrial Pgc1β −/− cardiomyocytes thus showed greater capacity to follow rapid pacing without producing pro-arrhythmic alternans phenomena. This would be compatible with their shorter APDs reflecting more rapid AP recoveries, despite their more prolonged activation processes.
Second, other arrhythmic syndromes have been contrastingly attributed to altered AP conduction, exemplified by murine Scn5a +/− hearts (Martin et al. 2012). This modifies the spatial extent and homogeneity of the travelling waves of AP excitation, or the quiescence that follows it. These were quantified by active wavelengths (λ) derived from conduction velocity [reflected by 1/(AP latency)] and APD terms (given by APD 90 ; Matthews et al. 2013), and resting wavelengths (λ 0 ) comprising DI 90 and AP latency terms (Matthews et al. 2013;Ning et al. 2016). Larger λ values reduce likelihoods that areas of depolarization and repolarization coincide at tissue heterogeneities, increasing the safety factor that ensures continued undisrupted propagation of the travelling wave (Weiss et al. 2005). Conversely, decreased λ values increase such likelihoods and those of the consequent wave break-ups into multiple wavelets, scroll-waves (Davidenko et al. 1995;Zaitsev et al. 2000;Pandit & Jalife, 2013) and further wavebreaks disrupting the AP conduction pathways (Spector, 2013). Clinical observations correspondingly associate short AP wavelengths with AF inducibility and maintenance (Hwang et al. 2015), particularly in AF patients (Padeletti et al. 1995).
The present analysis indeed demonstrated that in plots of λ against either BCL or λ 0 , Pgc1β −/− and, particularly, aged Pgc1β −/− atria consistently gave shorter λ in direct parallel with their pro-arrhythmic phenotype. Altered AP conduction with its consequences for its spatiotemporal properties thus constitutes a potential mechanism for the atrial arrhythmic changes associated with age and energetic compromise reported here.