Ageing in Pgc-1β^−/− mice modelling mitochondrial dysfunction induces differential expression of a range of genes regulating ventricular electrophysiology

Cardiac arrhythmias result when the normal sequence of electrical activation and recovery in the heart is disrupted. Of the many arrhythmic variants the ventricular arrhythmias, particularly ventricular tachycardia and ventricular fibrillation, constitute the primary cause of sudden cardiac death, a common cause of mortality and a significant public health burden especially in the ageing population [1,2]. Major risk factors for ventricular arrhythmia and sudden cardiac death include age [3] and metabolic disorders [4,5] such as obesity, diabetes and metabolic syndrome. Metabolic conditions themselves are associated with mitochondrial dysfunction which in turn is strongly linked to cardiac arrhythmias suggesting that impaired mitochondrial capacity constitute a further link between metabolic disorders and cardiac arrhythmias [6].

Pgc-1β is one of the co-activators of peroxisome proliferator activated receptor-γ (PPARγ) that plays a pivotal role in controlling genes that are responsible for the regulation of lipid and glucose metabolism [7]. PPARγ and Pgc-1β are highly expressed in white and brown adipose tissues and in other oxidative tissues including cardiac and skeletal muscles [8] where it is a key regulator of basal mitochondrial function maintenance [9]. Pgc-1’s pivotal involvement in the cellular energy balance means that modifications of these actors has been proven to serve as useful experimental models for the analysis of pathological changes associated with energetic deficiencies such as metabolic disorders [10].

To study the independent and combined effect of age and metabolic disorders on ventricular arrhythmic substrate, we have employed the PPARγ co-activator-1β (Pgc-1β) knockout mice model [11]. In our previous studies we have explored action potential (AP) and pro-arrhythmic characteristics in intact Langendorff-perfused hearts from both young and aged, wild-type (WT) and Pgc-1β^−/−mice [12–14]. We found that unprovoked young and aged Pgc-1β^−/− hearts subject to a pacing protocol showed increased incidences of alternans in both AP activation (upshoot velocity and latency), recovery duration and resting membrane potential (RMP) compared with WT hearts, and aged Pgc-1β^−/−hearts displayed increased incidence of actual arrhythmia. In conclusion Pgc-1β^−/− hearts show pro-arrhythmic instabilities attributable to altered AP conduction and activation [14] and Pgc-1β^−/−ventricular preparations showed consistently smaller Na⁺ currants independent of age even though other factors such as Na⁺ channel inactivation and K⁺ current activation were unchanged [15]. Moreover, morphometric assessments of the cardiac tissue indicated significantly increased fibrotic tissue with age and Pgc-1β^−/− genotype [15].

In the present study, our aim was to explore the transcriptional activities in the ventricular cardiac tissue relating to arrhythmic risks associated with ageing and mitochondrial energetic dysfunction. We have selected a set of genes known to regulate the physiological processes underlying excitable activity leading to arrhythmia [2] and other genes considered markers for fibrosis, inflammation and other pathological changes impacting on the action potential generation. The genes have been grouped according to their physiological relevance as presented below.

Na⁺–K⁺ ATPase α₁ and α₂ catalytic and accessory β₁ subunits (encoded by Atp1a1, Atp1a2 and Atp1b1) are related to the energetically dependent Na⁺–K⁺ ATPase-mediated membrane transport required to generate ionic gradients upon which the excitable activity is dependent.

The resting potential is mediated by background K⁺ channels. In this survey we have included a number of subtypes of ATP/ADP sensitive channels since we expect that the Pgc-1β^−/− genotype causes intracellular ATP deficiency and hence sensitivity of potassium conductance to ATP/ADP is highly relevant. Genes included are the ATP-sensitive inward rectifier K⁺ channel Kir2.2 (Kcnj12) which links resting potentials to intracellular nucleotide levels [16], the ATP-binding cassette (ABC) transporter subunits members 8 and 9 (Abcc8 and Abcc9), inwardly rectifying pore-forming K⁺ channels Kir6.1, Kir6.2 and Kir3.1 (Kcnj8, Kcnj11 and Kcnj3) and G protein-activated inward rectifier potassium channel 4, Kir3.4 (Kcnj5). In addition we have included a gene representing Cl⁻ channel conductance, Clcn3, encoding the chloride voltage-gated channel 3.

In respect of the Pgc-1β^−/− mice resting heart rate and their response to chronotropic activation with adrenergic challenge we have included the Hcn1, Hcn3 and Hcn4 isoforms of hyperpolarisation-activated cyclic nucleotide-gated channels mediating SAN pacemaker currents I_h and the α units of the voltage-dependent Na⁺ conductances, Scn5a and Scn7a, to survey the initiation of excitable activity.

To explore potential alterations in transcription of genes concerning the surface Ca²⁺ homeostasis we have included membrane voltage-dependent L-type Ca²⁺ channels Cav1.2 and Cav1.3 (Cacna1c and Cacna1d), T-type, Cav3.1 (Cacna1g) and Cav3.2 (Cacna1h), and the accessory β2 (Cacnb2), α₂/δ1 (Cacna2d1) and α₂/δ₂ subunits (Cacna2d2).

As well as surface Ca²⁺ homeostasis our survey covers molecules related to Ca⁺ homoeostasis between intracellular compartments. In this grouping we have included Atp2a2 that encodes one of the SERCA transporter proteins regulating the cytosolic Ca²⁺ reuptake into the sarcoplasmic reticulum (SR), the principal cardiac Na⁺–Ca²⁺ exchanger NCX (Slc8a1), the SR Ca²⁺ binding protein calsequestrin (Casq2) and the ryanodine receptor isoforms RyR2 (Ryr2) and RyR3 (Ryr3) responsible for the SR Ca²⁺ release.

Outward K⁺ currents regulate the action potential recovery phase that restores the transmembrane voltage to the resting potential level. These currents are mediated in murine hearts primarily by the voltage-sensitive transient outward current I_to carried by Kv1.4 and Kv4.3 (voltage-gated K⁺ channel subfamily A member 4 and subfamily D member 3, encoded by Kcna4 respectively Kcnd3) [2]. We also include Kv11.1, another voltage-gated K⁺ channel (subfamily H member 2), encoded by Kcnh2, mediating the rapid K⁺ current, IKr, in view of its clinical importance [17]. Moreover, assays to survey the expression of Kcnn1 and Kcnn2 encoding the Ca²⁺-activated K⁺ channels KCa2.1 and KCa2.2, are included in this group since they are thought to contribute to the action potential repolarisation phase particularly in atria however also in ventricular tissue [18]. Lastly, we assayed expression of Kcnk3 encoding the acid-sensitive K⁺ channel subfamily K member 3 and the regulatory KCNE1 subunit, encoded by Kcne11, in view of its association with some human Long QT syndromes [17].

The autonomous nervous system innervate the cardiac tissue and cause the action potential to be generated in the sinoatrial (SA) node and the impulse then travels into the ventricles via the atrioventricular node (AV node). To explore potential alterations in this system we have chosen both α and β adrenergic and muscarinic cholinergic receptors to represent a complete repertoire of sympathetic and parasympathetic cardiomyocyte receptors including the pivotal β₁ and β₂-adrenergic receptor subtypes (Adrb1 and Adrb2). In addition the α_1A, α_1B and α_1D adrenergic receptor subtypes (Adra1a, Adra1b and Adra1d), are included since they are believed to have protective roles against pathological remodelling in heart failure [19]. Chrm2, muscarinic acetylcholine receptor type 2 is assayed as the main acetylcholine responsive receptor balancing against the adrenergic receptors.

As a consequence of adrenergic stimulation the cAMP (cyclic adenosine monophosphate) pathway is activated leading to calcium influx followed by increase in muscle contractility, heart rate, electrical conduction velocity and relaxation rate. To capture potential alterations in this principal cell signalling pathway we have explored cardiac adenylyl cyclase, types 4 and 5 (Adcy4 and Adcy5), of which type 5 accentuates cardiomyopathic changes on chronic catecholamine stimulation, cGMP-dependent and cAMP-specific 3′,5′-cyclic phosphodiesterases 2A and 4D (Pde2a and Pde4d), the protein kinase A catalytic α-subunit (Prkaca), genes encoding the cAMP-dependent protein kinase regulatory subunits type I-α, II-α and II-β (Prkar1a, Prkar2a and Prkar2b) [20] and finally, we assay the major cardiac, calcium/calmodulin-dependent protein kinase, type II-δ (Camk2d), which mediates numerous cellular responses to Ca²⁺ signals [21].

In our previous physiological studies we have reported increased fibrotic change with age in mice with the Pgc-1β⁻/⁻ genotype [22] and cardiac fibrosis is generally associated with altered cardiac conduction, (slowing, block and re-entry) [23]. In this study we examine fibrotic markers including the cytokine transforming growth factor β1 (TGF-β1; Tgfb1) [24], the gap junction forming protein δ 3, also known as Connexin 30.2 in mouse, encoded by Gjd3, the collagen precursor Col3a1 encoding the collagen type III α1 chain and Col1a1 encoding the major component of type I collagen, the fibrillar collagen found in most connective tissues [25].

The final group of genes tested are thought to act as markers for a range of developmental, inflammatory and hypertrophic changes ultimately impacting the AP generation and propagation. The transcriptional repressor Tbx3, encoded by Tbx3, is known to affect particular components of the cardiac conduction system, [26]. Myh6 encodes the major thick filament protein MHC-α which is associated with late-onset hypertrophic cardiomyopathy and sinus node disorder if carrying mutations [2]. The cardiac hormone ANP (atrial natriuretic peptide), encoded by Nppa, is a known biomarker for heart disease, specifically, deficiency causes hypertension and cardiac hypertrophy which is related to its function as a regulator of the sodium balance. However recent studies has provided evidence that ANP also has a role in regulating vascular remodelling and energy metabolism [27]. The non-specific ion channel TRPC1 (Trpc1) conducts both Ca²⁺ and Na⁺ and has been found to be an important regulator of cardiac hypertrophy [28]. Finally in this miscellaneous group we have included genes encoding Pgc-1β itself and the complementary Pgc-1α (Ppargc1b and Ppargc1a).

Experiments were approved by the University of Cambridge ethics review board under a U.K. project license for studies of cardiac arrhythmia. All procedures complied with the U.K. Home Office regulations (Animals (Scientific Procedures) Act 1986). The experiments also conformed to the Guide for the Care and Use of Laboratory Animals, U.S. National Institutes of Health (NIH Publication Number 85-23, revised 1996). Mice were housed in an animal facility at 21°C with 12-h light/dark cycles. Animals were fed sterile chow (RM3 Maintenance Diet, SDS, Witham, Essex, U.K.) and had free access to water, bedding and environmental stimuli. Mice were killed by cervical dislocation. No recovery, anesthetic or surgical procedures were required. WT C57/B6 and Pgc-1β^−/− (The Jackson Laboratory, Maine, U.S.A.) mice were bred for the experimental protocols. Mice were bred on a C57/B6 background to avoid possible strain-related confounds.

The mice were divided into four groups: young WT (n=3), young Pgc-1β^−/− (n=3), aged WT (n=3) and aged Pgc-1β^−/− (n=3). Young mice were between the ages of 12 and 16 weeks and aged mice greater than 52 weeks. After extraction, the heart was cut to separate the ventricular tissue from the rest of the heart and subsequently snap-frozen.

RNA was extracted from fresh frozen tissues, stored in −80°C, with the Qiagen RNeasy mini Plus kit. Cardiac ventricular tissue were weighed and quickly minced on ice. A third of the tissue, approximately 30 mg, was used in the next step of the RNA isolation protocol. The tissue was homogenised in RLT buffer supplemented with β-mercaptoethanol with a Stuart handheld homogeniser until completely smooth. Genomic DNA was eliminated by centrifugation through a column supplied with the kit prior to extraction of the RNA according to the manufacturer’s protocol. RNA integrity was assessed by using an Agilent Bioanalyzer to obtain RNA integrity numbers (RIN) according to the manufacturer’s protocol. RNA samples with RINs above 8 were used for the study. The RNA was used to prepare cDNA with High Capacity cDNA Reverse Transcription Kit (Applied Biosystems) according to the manufacturer’s instructions.

The efficiency of the cDNA protocol was tested by preparing the cDNA from a serial dilution of the RNA and then these cDNA samples were run with a SYBRgreen qPCR protocol confirming equal efficiency over a range of RNA concentrations. cDNA was also confirmed negative for genomic DNA contamination.

Each custom-made card contained 64 pre-validated assays in triplicate with a reaction volume of 1 μl. The cards were run exactly according to instructions specific for the cards. Briefly, the cDNA (100 ng/well) was mixed with 2× Mastermix from Thermo Fisher, 100 μl was loaded in each well slot on the cards, the cards were then spun down and sealed and run on a Quant 7 cycler. The amplification conditions were: 50°C for 2 min and 95°C for 10 min for the initial DNA melting and inactivation of the reverse transcription reaction, followed by 40 cycles of 95°C for 15 s and 60°C for 60 s.

Analysis of the Taqman array card data was performed by using the Quant studio software and Microsoft Excel by calculating fold changes with the delta-delta-C_T method [29]. The threshold was set at 0.2 fluorescence units and the baseline range was set to automatic assignment. The mean of the Cq values for the genes HPRT, Gapdh and ActinB were used as references.

P-values were calculated with Student’s independent ttests and Analysis of variance (Anova, one-way and two-way) and Tukey’s Honest Significant Difference (HSD) test where appropriate. Prior to ttests and Anovas the homogeneity of variance for the sample groups were tested with Levene’s test and were found negative (equal variance) for all groups. Statistical analyses were performed in the R programming language and statistical significance was assumed at P<0.05. Since our data to a high degree are exploratory with an aim to discover genes with differential expression we have opted to present the unadjusted P-values rather than with correction for multiple testing. We do acknowledge the risk of false positives and confirm that adjusting the P-values results in no statistical significance.

In the present study our focus has been to explore the effects on gene expression of both Pgc-1β deficiency and age at the individual gene level and as functional groups. To be able to analyse a large number of genes we used Custom Taqman array cards that were pre-probed with 60 cardiac function relevant genes as detailed above. The samples were divided into four groups: WT Young (n=3), WT Old (n=3), Pgc-1β^−/− Young (n=3) and Pgc-1β^−/− Old (n=3).

To get a detailed overview of the differential expression of all tested genes Figure 1 present the raw expression fold changes as a heatmap with accompanying Anova single group and main effect significance statistics. The heatmap visualises the mean fold-changes for each tested group (Young-WT, Young-Pgc-1β^−/−, Old-WT and Old-Pgc-1β^−/−) at the individual gene level. At the top of the heatmap are the genes with the largest average decrease (green) in expression compared with the level in WT-Young samples (yellow) and at the bottom are the genes with the highest average increase in expression (red). As evident from the heatmap the majority of the investigated genes are relatively unchanged with both age and Pgc-1β deficiency, fold changes range from 0.38 to 2.06 compared with the WT-Young group. Two-way Anova main effect analyses however show a handful of significantly differentially expressed genes. The Pgc-1β^−/− genotype significantly decrease the expression level of the ATPase activity regulating genes Atp1a1 and Atp1b1, the adrenoceptor Adra1d, the potassium channels Kcnj8 and Kcnk3, and Casq2 encoding the calcium-binding protein Calsequestrin. The age factor significantly affects the level of the fibrotic marker Col3a1, potassium channels Kcnj5 and Kcnj3, the pacemaker Hcn3, and the adrenoceptor Adra1b. Interestingly, all, but the fibrotic marker, are found to be up-regulated with age (Figure 1).

In addition, the heatmap can be employed to distinguish genes where the effects of age and genotype are not necessarily substantially different however follow the same direction in the sense that the effect of age and genotype are potentially additive. We found five genes (Nppa, Col3a1, Kcnn2, Trpc1 and Col1a1) that were consistently down-regulated with age and Pgc-1β^−/− genotype and we found two genes (Hcn4 and Gjd3) that were consistently up-regulated with these factors (Figure 1). To further identify and analyse dependent interactions between age and Pgc-1β deficiency we also analysed the ‘between subject effects’ of the Two-Way Anova for age and Pgc-1β^−/−. Only one gene, Hcn3, was identified as showing a significant (P=0.023) dependent interaction for the two tested factors. In this case that translates to that the expression change relating to age is dependent on the genotype and vice versa. For the other genes we found no significant dependencies between age and genotype nonetheless excluding the discussed general additive effects.

One-way Anova comparing the four groups (Young-WT, Young- Pgc-1β^−/−, Old-WT and Old-Pgc-1β^−/−) indicate significantly differentiating expression in only three genes (Atp1b1, Kcnj3 and Hcn3) with Tukey’s HSD post hoc tests only pinpointing the pacemaker gene Hcn3 in Old-Pgc-1β^−/− as being significantly different from the other groups (Old-Pgc-1β^−/− vs Young- Pgc-1β^−/−; P=0.026 and Old-Pgc-1β^−/− vs Old-WT; P=0.045) (Figure 1).

To visualise and analyse the separate effects of age and genotype volcano plots displaying the magnitude of change against the level of statistical significance were produced by plotting the unadjusted P-values from pairwise-independent Student’s t tests against the log2 transformed expression fold changes for the Pgc-1β^−/− in young respectively Old mice (Figure 2A) and for the Old in WT respectively Pgc-1β^−/− mice (Figure 2B).

Overall, the volcano plots confirm the heatmap analysis that the expression differences among the groups are relatively small and limited to a handful of genes. Even so, comparing WT with Pgc-1β^−/− genotype in Young and Old mice separately clearly display that the majority of studied genes are less expressed in the Pgc-1β deficient mouse model in both Young and Old mice (Figure 2A). In young mice 54 out of 59 genes (not including Ppargc1b) were less expressed with an average decrease of 22%. In old mice the difference was less clear with 43 out of 59 genes less expressed with an average decrease of 9%. On single gene level five genes in the aged mice (Atp1b1, Scn7a, Casq2, Kcnk3 and Adra1b) and none in the young mice were found to have significantly reduced gene expression in the Pgc-1β^−/− genotype compared with normal as indicated in Figure 2A. Three of these genes, Atp1b1, Casq2 and Kcnk3, were the same as found to be statistically different when assessed with Anova main effect analysis as well (Figure 1). Interestingly, when separating according to age it is evident that the statistical difference is bearing on the Old mice group and not found in Young mice. We also observed that the gene expression of the sodium ion channel encoded by Scn7a and the α-1 adrenergic receptor (Adra1b) were both significantly decreased in the aged mice however not in the young mouse population (Figure 2A).

In Figure 2B the expression levels of young and old mice in WT respectively Pgc-1β deficient mice are displayed in the same format as Figure 2A. Looking at these volcano plots it is evident that comparing tissue from young mice with aged mice (Figure 2B) show a significant increase in expression in aged mice, especially in the Pgc-1β knockout mice, where 45 out of 59 genes were more expressed with an average increase of 34%. In the WT group the difference was smaller with 40 out of 60 genes more expressed in old compared with young mice with an average increase of 11%. Despite the broadly increased expression levels in aged mice only three genes (Kcnj3, Kcnj5 and Hcn3) in the Pgc-1β^−/− mice, and none in the WT mice, were found to be significantly increased. As discussed above the Kcnj3 and Kcnj5 encodes potassium channel proteins important for the maintenance of the resting membrane potential and Hcn3 is one of the hyperpolarisation-activated cyclic nucleotide-gated (HCN) channels that are vital for normal pace making.

In conclusion, the heatmap (Figure 1) and the volcano plots (Figure 2A) demonstrate that on a single gene level, there are a handful of genes that are significantly differentially expressed when comparing Pgc-1β deficient mice with WT in young and old mice. Even though most genes only show small changes in fold expression it is clear that knocking out Pgc-1β has a substantial effect on gene expression with an altogether significant decrease in expression of the studied set of genes. On the other hand, our data show that comparing the gene expression in young mice with old mice in both WT and mice with compromised mitochondrial function there is a general increase in gene expression with age in this particular set of genes (Figures 1 and 2B).

In the present study we have found that mitochondrial dysfunction and age both have relatively limited impact on the gene expression of our tested set of genes relevant for the cardiac action potential. At single gene level we have found a handful genes to be differentially expressed to a significance level P<0.05 as presented in Figures 1 and 2. To further explore the changes in gene expression pattern and potentially correlate these changes with the functional changes seen on the electrophysiological level [14,15,22] we have subgrouped the genes into ten groups according to their functions as indicated in Table 1 and discussed in the introduction.

Two-way Anova main effect statistics for age and genotype show that four of the functional groups are significantly down-regulated comparing WT with Pgc-1β^−/− genotype and three of the groups are significantly up-regulated comparing Young with Old mice.

As a group the genes relating to the Na/K ATPase activity show a small but clear decrease in the Pgc-1β deficient group compared with WT in both young and old mice (Table 1). Also the ion channels relating to the resting membrane potential are, as a group, down-regulated in the Pgc-1β^−/− mice. On the contrary, comparing young with old mice there is a significant increase in gene expression of these genes again indicating changes of the resting potential maintenance however the genotypic changes are not additive to the effects of ageing. Exploring the genes relevant for the calcium homoeostasis we found only very minor changes. As presented in Figure 2A, the Casq2 gene is slightly less expressed in old Pgc-1β knockout mice compared with WT. Genes regulating the calcium surface homoeostasis are not differentially expressed on the single gene level however as a group there is a significant increase with age. The same pattern was found when exploring the expression of the genes responsible for the repolarisation phase, at a single gene level only one gene showed a significant change in expression however as a group there is a statistical decrease in the Pgc-1β^−/− genotype and also a distinct gene expression increase with age (Table 1). Adrenergic receptors mediating the sympathetic signalling in the cardiac tissue was found to be mainly down-regulated in our knockout model but then significantly increased as a group with age. The cholinergic receptor Chrm2 showed the same pattern seemingly counteracting the adrenergic effects (the Anova group analysis is performed without the Chrm2 gene expression). Following the gene expression changes of the adrenergic receptors the genes constituting the cAMP pathway are as a group significantly down-regulated with Pgc-1β deficiency and up-regulated with age even though at singular gene level the differences are considered non-significant. Lastly, studying our fibrotic markers as a group did not display any altered gene expression indicating that the decrease in the Col3a1 with age discussed above might not be a general indicator of fibrotic change.

We have performed an extensive transcriptional profile of genes encoding molecules related to excitable activity (depolarisation, resting potential maintenance and repolarisation), calcium homoeostasis and fibrotic change which have been implicated in cardiac arrhythmias. Overall our aim has been to compare and relate the Pgc-1β^−/− genotype and age-dependent phenotype with its transcriptional profile specifically regarding arrhythmic substrate, reductions in conduction velocity reflecting reduced Na⁺ current and increased tissue resistance.

The Pgc-1β^−/− mouse model has proven to be a relevant model to study ‘metabolism induced’ arrhythmia [10,30]. Several previous studies [15,22,31,32] have explored the effects that this mitochondrial dysfunction has on the ventricular heart and it is now well established that mice deficient in Pgc-1β present with an age-dependent pro-arrhythmic phenotype. In the present study our aim was to investigate the molecular mechanisms causing this phenotype. We here describe how a set of genes known to determine the specifics of the AP and the cardiac conduction, including ion channels, fibrotic markers, cAMP pathway members and others, together form a complex pattern of regulation. We found that despite the clear physiological effects, the actual differences on gene level are relatively small. However, even though on single gene level the differential expression is minimal, we do find that grouping genes according to function shows a pattern of regulation. Interestingly, we found that, particularly in young mice, a vast majority of the studied genes are down-regulated in ventricular cardiac tissue from the Pgc-1β^−/− mice. A natural explanation for this could be that since Pgc-1β deficiency causes the mitochondria to dysfunction and thereby lead to a lack of energy in the cell which in turn can cause lesser gene expression on the molecular level and arrhythmia on the functional level. Surprisingly though, with age, this same set of genes are mainly up-regulated, especially in the Pgc-1β deficient mice. It is possible that this up-regulation of gene expression is a compensation mechanism developed with age, however unlikely via Pgc-1α regulation since the level is unchanged across the samples.

We here suggest that the electrophysiological effects we have observed in the Pgc-1β deficient mouse model previously is potentially due to changes in gene expression of genes encoding proteins bearing on the Na–K ATPase activity (specifically Atp1a1 and Atp1b1), the resting membrane potential (specifically Kcnj3, 5 and 8), repolarisation (specifically Kcnk3), the calcium balance (specifically Casq2) and proteins mediating the cyclic nucleotide signalling in response to autonomic activation. These groups of genes are to a high degree also involved in the changes seen with age pointing to that there might be overlapping mechanisms regarding ventricular arrhythmia in relation to metabolic disease and old age.

Characterisation of the action potential parameters by means of intact Langendorff-perfused hearts clearly point to increased pro-arrhythmic substrate in both older hearts and hearts with mitochondrial energetic deficiency [14,22]. For example the incidence of AP latency alternans was found to increase with age and Pgc-1β^−/− genotype and the general frequency of actual arrhythmia was measured to be significantly higher in the aged Pgc-1β^−/− hearts. However, even though initial ECG recordings showed shorter ventricular effective refractory periods (VERPs) for both young and aged Pgc-1β^−/− hearts with the aged Pgc-1β^−/− heart markedly the shortest the following incremental pacing protocols indicated that the young, not old, Pgc-1β^−/− hearts were least able to tolerate the shortened basic cycle length (BCL) at which hearts failed the pacing rate and entered block or ventricular arrhythmia. Moreover, both action potential duration (APD90) and the diastolic interval (DI90) was observed to be unchanged with age in the WT hearts however the APD90 was decreased in young Pgc-1β^−/− hearts and increased in old Pgc-1β^−/− hearts and the DI90 was increased in the former and decreased in the latter compared with the level in the WT hearts [14,22]. Furthermore, regular 8 Hz pacing was conducted to resemble normal resting heart rate and action potential properties were characterised. Clear differences in maximum cardiac upstroke rate ((dV/dt)max) and AP duration were demonstrated. Interestingly both these factors were significantly dropped in Young Pgc-1β^−/− compared with young and old WT hearts however in the old Pgc-1β^−/− levels were picked up and even increased compared with WT hearts. The regular 8 Hz pacing did not cause spontaneous arrhythmic events in any of the experimental groups however with programmed electrical stimulation protocol a range of abnormal rhythms were elicited. Of particular interest was the counts of non-sustained ventricular tachycardia (NSVT) and sustained ventricular tachycardia (SVT). The latter was only recorded in the aged Pgc-1β^−/− however NSVT was significantly most frequent in the young Pgc-1β^−/− hearts [22].

Our aim with this transcriptional profiling study was to relate the physiological phenotypes of Pgc-1β^−/− and old age with the gene expression of targeted genes and our initial expectation was that mitochondrial dysfunction as a model for metabolic disease and ageing would be factors that independently and additively both increased the pro-arrhythmic substrate. Our data however indicate that the incidence of pro-arrhythmic substrate along with the gene expression follows a more complex pattern than anticipated. On a gene expression level the present data demonstrate that most genes are unchanged or changed only to a small degree and that age and mitochondrial dysfunction does not necessarily constitute additive effects. It is plausible that the effect of ageing and metabolic disorders, though similar, differs mechanistically on the molecular level. Interestingly, on a physiological level our previous data establish that ageing respectively metabolic disorder are not always mutually making symptoms worse indicating that there is indeed some congruence between gene expression and phenotype.

Considering the previously reported decrease in Na⁺ current and slowed conduction velocity we were surprised to find only the NaV2.1 (encoded by Scn7a) and not the main voltage-gated sodium channel NaV1.5 (Scn5a) to be down-regulated with age and mitochondrial dysfunction. Equally surprising we found no strong evidence that the previously described fibrotic activity was directly caused by transcriptional regulation. The two tested collagens in our study, Col1a1 and Col3a1, are both marginally down-regulated, while the Gap Junction δ 2 protein, encoded by Gjd3, is up-regulated which at most suggest an imbalance in the fibrillar collagen regulation.

In conclusion we here provide evidence that the observed changes in the cardiac electrophysiology characteristics following ageing and mitochondrial dysfunction might be caused by a complex of surprisingly small transcriptional gene expression alterations. It is also possible that these observed changes are a consequence rather than a cause of the physiological changes and other yet unknown mechanisms control the pro-arrhythmic substrate.

This work was supported by the Medical Research Council [grant number MR/M001288/1]; the Wellcome Trust [grant number 105727/Z/14/Z]; the British Heart Foundation [grant numbers PG/14/79/31102, PG/15/12/31280]; the Sudden Adult Death Syndrome (SADS) U.K. Society; and the Fundamental Research Grant Scheme [grant number FRGS/2014/SKK01/PERDANA/02/1] (Ministry of Education, Malaysia).

C.E.E., I.T.F. and K.R.C. performed the genome transcription studies. S.A., H.V. and A.A.G. performed specimen preparation. C.L.-H.H. and K.J. conceived and designed the study. C.E.E., C.L.-H.H. and K.J. analysed the data and wrote the manuscript.

The authors declare that there are no competing interests associated with the manuscript.

Anova

analysis of variance

ANP

atrial natriuretic peptide

HSD

honest significant difference

Pgc-1β

promoter peroxisome proliferator activated receptor-γ co-activator-1β

PPARγ

peroxisome proliferator activated receptor-γ

RIN

RNA integrity number

SERCA

sarcoendoplasmic reticulum calcium transport ATPase

SR

sarcoplasmic reticulum

WT

wild-type