Atrioventricular node dysfunction in pressure overload-induced heart failure—Involvement of the immune system and transcriptomic remodelling

Heart failure is associated with atrioventricular (AV) node dysfunction, and AV node dysfunction in the setting of heart failure is associated with an increased risk of mortality and heart failure hospitalisation. This study aims to understand the causes of AV node dysfunction in heart failure by studying changes in the whole nodal transcriptome. The mouse transverse aortic constriction model of pressure overload-induced heart failure was studied; functional changes were assessed using electrocardiography and echocardiography and the transcriptome of the AV node was quantified using RNAseq. Heart failure was associated with a significant increase in the PR interval, indicating a slowing of AV node conduction and AV node dysfunction, and significant changes in 3,077 transcripts (5.6% of the transcriptome). Many systems were affected: transcripts supporting AV node conduction were downregulated and there were changes in transcripts identified by GWAS as determinants of the PR interval. In addition, there was evidence of remodelling of the sarcomere, a shift from fatty acid to glucose metabolism, remodelling of the extracellular matrix, and remodelling of the transcription and translation machinery. There was evidence of the causes of this widespread remodelling of the AV node: evidence of dysregulation of multiple intracellular signalling pathways, dysregulation of 109 protein kinases and 148 transcription factors, and an immune response with a proliferation of neutrophils, monocytes, macrophages and B lymphocytes and a dysregulation of 40 cytokines. In conclusion, inflammation and a widespread transcriptional remodelling of the AV node underlies AV node dysfunction in heart failure.


Statistical analysis of gene expression
For RNA-Seq data, P values were corrected for multiple testing using the Benjamini-Hochberg method to control the false discovery rate. In figures, data are shown as mean±SEM. A P value of <0.05 was regarded as significant, with asterisks indicating significance on figures. Hierarchical clustering was performed using Pearson's correlation distance and ward.D2 agglomeration method and a heatmap was generated ( Figure 2C) for the 3,077 differentially expressed transcripts (adjusted P value <0.05) using ComplexHeatmap v2.2.0 (https://www.bioconductor.org/packages/release/bioc/html/ComplexHeatmap.html). The Z-score is the number of standard deviations a given data point lies above or below the mean.

Gene ontology (GO) enrichment analysis
Enrichment analysis was performed using the R package topGO v2.38.1 (Bioconductor; https://bioconductor.org/packages/release/bioc/html/topGO.html). Analysis was performed on downregulated and upregulated differentially expressed genes identified using RNAseq. Significance was tested with Fisher's exact test and the elim algorithm. Significantly enriched biological processes related to cardiac function and HF were selected.

Qiagen Ingenuity Pathway Analysis (IPA)
IPA (https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis) was used for canonical pathway analysis. All significantly expressed transcripts were entered into IPA and core analysis was performed. Default parameters were used for the analysis. Right-tailed Fisher's exact test was performed to assess the statistical significance of enriched pathways. P<0.05 was considered significant.

Further information on electrophysiological changes, upregulation of HF markers, and remodelling of ion channel and related transcripts in HF
The heart rate measured in the conscious mouse using an ECGenie steadily declined following TAC surgery (Figures 1F and S1G) as a result of sinus node dysfunction (Yanni et al., 2020). There was also a significant increase in the QRS interval indicative of His-Purkinje dysfunction (a slowing of conduction through the His-Purkinje system) and uncorrected and corrected QT intervals indicative of an increase in ventricular action potential duration by the end of the experiment in the HF mice ( Figure 1H-J).
It is well known that Nppa (responsible for atrial natriuretic peptide, ANP), Nppb (responsible for brain natriuretic peptide, BNP) and Myh7 (responsible for b-myosin heavy chain) are upregulated in HF (Man et al., 2018;Dirkx et al., 2013) and all three transcripts were significantly upregulated in the AV node in HF ( Figure S2).
Knockout of Hcn4 results in AV block (Baruscotti et al., 2011). There was a ~25% downregulation of Hcn4 in HF, but it was not significant ( Figure S4A). Seven Clcn transcripts for Clchannels were detected (Clcn4>Clcn7>Clcn3>Clcn6>Clcn1>Clcn5>Clcn2) and of these Clcn1 and Clcn2 were significantly downregulated ( Figure S4B). The downregulation of Clcn2 may be important, because the CLCN2 channel carries slowly activating inward current at diastolic potentials like the HCN4 channel and has been shown to contribute to pacemaking in the sinus node (Huang et al., 2009). The downregulation of Clcn1 may also be important, because myotonic dystrophy patients have abnormal splicing of Clcn1 and an increased incidence of a prolonged PR interval and AV block (McNally and Sparano, 2011) (although the abnormal Clcn1 splicing may only be associated with the AV node dysfunction and not the cause of it).
Inhibition of the 'Ca 2+ clock' can slow AV node conduction and increase the PR interval (Saeed et al., 2018). As already discussed, some Ca 2+ channel transcripts were downregulated. However, other Ca 2+ clock transcripts were ether unaffected or upregulated; for example, Slc8a1 and Casq2, responsible for the Na + -Ca 2+ exchanger and calsequestrin 2, were both significantly upregulated ( Figure S5). Via the Na + -Ca 2+ exchanger, the intracellular Ca 2+ concentration is closely linked to the intracellular Na + concentration, which is set by the Na + -K + pump. In HF, there was a significant upregulation of Atp1a1 (a1 isoform of the Na + -K + pump) and a significant downregulation of Atp1a2 (a2 isoform) ( Figure S6). In the heart, the a2 isoform preferentially assembles with the b2 isoform (Atp1b2) (Clausen et al., 2017) and in HF Atp1b2 was downregulated by a similar percentage as Atp1a2 ( Figure S6). The a2 isoform localises close to the Na + -Ca 2+ exchanger and may help in the regulation of intracellular Ca 2+ (Clausen et al., 2017). The FXYD family including phospholemman (FXYD1) has been demonstrated to regulate the Na + -K + pump; two of the Fxyd transcripts showed significant changes ( Figure S6C).
Gap junctions made up of connexins are key determinants of action potential conduction, but no changes in connexin transcripts known to be linked to AV node conduction were detected ( Figure  S7).

Evidence of activation of multiple intracellular signalling pathways and transcription factors in HF
Protein kinase A. The role of the sympathetic nervous system and the b-adrenergic receptor pathway in causing or exacerbating cardiac disease has long been recognised (Bernstein et al., 2011). b-adrenergic receptor stimulation leads to an activation of protein kinase A via adenylate cyclase and a G protein. In the AV node in HF, there was no change in transcripts for b-receptors, the G protein or adenylate cyclase (see below). Protein kinase A is made up of catalytic and regulatory subunits and whereas catalytic subunit transcripts were unaffected, regulatory subunit transcripts were affected; Prkar1a was the most abundant of the regulatory subunit transcripts and it was upregulated by 76% ( Figure S10A). A similar pattern of change in protein kinase A subunits (protein) has been observed in HF in the human (Han et al., 2013). Phosphodiesterases able to hydrolyse cAMP (Figure S10B,C) antagonise protein kinase A by hydrolysing cAMP. In the AV node in HF, the expression of the most abundant of the cAMP-selective phosphodiesterase transcripts, Pde4a, was downregulated ( Figure S10C). Ablation of Pde4 has been shown to affect excitationcontraction coupling and predispose to the development of HF; downregulation of Pde4a is observed in the failing human heart (Richter et al., 2011). Expression of some transcripts for other phosphodiesterases able to hydrolyse cAMP were also significantly changed or showed a trend towards a change in HF (Figure S10B,C). In contrast, there were no changes in cGMP-selective phosphodiesterase transcripts ( Figure S10D).
Ca 2+ -calmodulin-dependent protein kinase II (CaMKII). CAMK2D is upregulated in HF, and transgenic mice overexpressing Camk2d develop a dilated cardiomyopathy (Zhang et al., 2003). Sinus node dysfunction in HF has been attributed to activation of CaMKII (Swaminathan et al., 2011). Transcript for the dominant CaMKII isoform, Camk2d, was significantly upregulated in the AV node in HF ( Figure S11).
Hippo pathway. Studies have demonstrated a role for the Hippo pathway in cardiac disease (Ikeda et al., 2019;Chen et al., 2020). On stimulation, MST1/2 and its adaptor protein, SAV1, are phosphorylated and activated, and they in turn phosphorylate and activate the LATS1/2-MOB1 complex ( Figure S12A). The activated LATS1/2 phosphorylates YAP1 leading to the cytoplasmic retention and possible degradation of the YAP1/TAZ complex ( Figure S12A) (Chen et al., 2020). When the Hippo pathway is inactive, YAP1/TAZ is mainly localised in the nucleus and together with transcriptional partners (e.g. TEADS) initiates or impedes transcription of target genes ( Figure S12A) (Chen et al., 2020). The Hippo pathway can be cardioprotective in HF, although it can also exacerbate it (Chen et al., 2020). Clinical ischaemic heart disease and idiopathic dilated cardiomyopathy is characterised by increased YAP1/TAZ protein levels and transcriptional activity and consequent upregulation of target genes such as Ccn2 (responsible for connective tissue growth factor, CTGF) (Chen et al., 2020). CTGF is a central mediator of tissue remodelling and fibrosis and its inhibition can reverse the process of fibrosis (Lipson et al., 2012). Inactivation of CTGF (using a monoclonal antibody) in mice following myocardial infarction reduced the heart weight:body weight ratio, left ventricular mass, cardiomyocyte hypertrophy, and fibrosis (Vainio et al., 2019). There were many changes to the Hippo pathway in the AV node in HF ( Figure S12B). Many changes potentially favour nuclear localisation of YAP1/TAZ. In addition, Tead1, Tead4 and Wwtr1, necessary for YAP target gene transcription, were upregulated. One of the target genes at least, Ccn2 (CTGF), was upregulated ( Figure S12B).
WNT signalling. Although WNT-signalling is quiescent under normal conditions, it is activated by pathological stress, and activation of WNT signalling is sufficient for the induction of cardiac hypertrophy and cardiomyopathy (Malekar et al., 2010;Zhao et al., 2018;Foulquier et al., 2018). WNT protein binds to a Frizzled receptor; co-receptors (LRP5, LRP6, ROR1 and ROR2) may be required (Foulquier et al., 2018). This causes an accumulation of β-catenin in the cytoplasm and its eventual translocation into the nucleus to act as a transcriptional coactivator of transcription factors. Various Wnt transcripts (including the most abundant, Wnt9b), various Fzd transcripts responsible for Frizzled receptors (including the most abundant, Fzd4), and Ctnnb1 (responsible for β-catenin) were significantly upregulated in HF ( Figure S13).
Protein kinases and the protein phosphatase interactome. Protein phosphorylation is an important signalling mechanism, and it is determined by a balance of protein kinases and phosphatases. Some kinases have already been considered; in total there were significant changes in 109 protein kinase transcripts ( Figure S14). Some phosphatases (DUSPs) have again already been considered. Protein phosphatase type-1 (PP1) plays an important role in cardiac physiology and pathophysiology (Chiang et al., 2016;Chiang et al., 2018). PP1 is known to play an important role in HF and some studies have reported the activity of PP1 to be increased in HF (presumably all what is known relates to the ventricles) (Chiang et al., 2016). PP1 is made up of a catalytic subunit and regulatory subunits. Transcript for the catalytic subunit has been reported to be upregulated in HF patients (Chiang et al., 2016). Two regulatory subunits, PPP1R3A (Cordero et al., 2019) andPPP1R7 (Chiang et al., 2018), have been implicated in HF; mice lacking PPP1R3A are protected against HF (TAC model) (Cordero et al., 2019). In this study, transcript for one of two abundant catalytic subunit isoforms, Ppp1cb, was significantly upregulated in the AV node in HF ( Figure S15). There was a significant upregulation of the regulatory subunit Ppp1r3a, but not of Ppp1r7 ( Figure  S15). There was a significant up and down regulation of transcripts for other regulatory PP1 subunits ( Figure S15). The largest upregulation was of Ppp1r3c ( Figure S15). PPP1R3C is involved in the control of metabolism: during hypoxia, hypoxia-inducible factor 1 (HIF1) promotes glycogen accumulation by regulating PPP1R3C (Shen et al., 2010). In this study, Hif1a, as well as Ppp1r3c, was significantly upregulated in the AV node in HF ( Figure S15). In this study, another regulatory subunit, Ppp1r1a, was significantly downregulated in HF ( Figure S15). This is significant, because PPP1R1A is a potent inhibitor of the catalytic subunit of PP1 and a downregulation of PPP1R1A in HF has been observed in other studies (Chiang et al., 2016). In the present study, in HF, an upregulation of various subunits for protein phosphatases types 2, 3 and 4 (PP2-4) was also observed ( Figure S15). PP2 is a critical regulatory molecule in both health and disease, with a myriad of targets in heart (DeGrande et al., 2013;Lubbers and Mohler, 2016). There can be an increase in PP2 in HF, including in the human, and modulation of PP2 may contribute to the pathophysiology of cardiac disease (DeGrande et al., 2013;Lubbers and Mohler, 2016).
Other pathways. In addition to the pathways above, there was evidence of changes in the JAK-STAT pathway (considered below) and the NOTCH pathway (significant downregulation of Notch2 to 85% of control and significant upregulation of Notch4 to 173% of control).

Remodelling of further extracellular matrix transcripts in HF
Glycoproteins and proteoglycans. Glycoproteins make the extracellular matrix a cohesive network of molecules (Megías et al., 2019). They link structural molecules to each other, and also to cells (Megías et al., 2019). Fibronectins, laminins and tenascins are major glycoproteins (Megías et al., 2019). Fibronectin (encoded by Fn1) binds integrins in the cell membrane to components of the extracellular matrix such as collagen and fibrin (Megías et al., 2019); Fn1 was significantly upregulated in HF ( Figure S17A). Laminins are cell adhesion molecules found predominantly in basement membranes (Megías et al., 2019); one of their functions is to interact with receptors in the cell membrane and thereby regulate signalling pathways (Megías et al., 2019); there were significant changes in three laminin subunit transcripts in HF ( Figure S17A). Tenascin-C (Tnc) is a large extracellular matrix glycoprotein, which is upregulated during physiological and pathological remodelling and is involved in important signalling pathways (Imanaka-Yoshida et al., 2020); Tnc was significantly upregulated in the AV node in HF ( Figure S17A). Elastin (encoded by Eln) is a glycoprotein and provides elasticity unlike the collagens (Wang et al., 2021a) and Eln was significantly upregulated in HF ( Figure S17A); will this increase the elasticity of the sinus node? Proteoglycans are a major component of the extracellular matrix and form the "filler" substance between cells (https://en.wikipedia.org/wiki/Proteoglycan), but only two proteoglycan transcripts Bgn and Chadl were significantly affected in HF ( Figure S17A). Biglycan (encoded by Bgn) is related to inflammation: upregulation in adipose tissue during inflammation may be involved in the perpetuation of the inflammatory milieu (Adapala et al., 2012).
Extracellular matrix affiliated. There are 165 'extracellular matrix affiliated' transcripts listed in Matrisome and there were 39 significant changes ( Figure S17B).
Secreted factors. There are 367 'secreted factors' listed in Matrisome and these include signalling molecules known to affect the extracellular matrix; there were 61 significant changes ( Figure S18A). TGF-b and WNT signalling are known to play an important role in cardiac fibrosis (Yousefi et al., 2020) and Tgfb1, Tgfb2, Wnt4, Wnt7a and Wnt9b were all upregulated as already discussed ( Figures 7C and S18A).
Extracellular matrix regulators. There are 304 extracellular matrix regulators listed in Matrisome and these include the metalloproteinases (Mmp) and ADAMTS proteases (Adamts) for example (Santamaria and de Groot, 2020); there were 55 significant changes ( Figure S18B).

Further evidence of an immune response in HF
A large number of cytokine transcripts were significantly altered in the AV node in HF ( Figure  7C). Two cytokines, IL-1b (interleukin 1 beta; Il1b) and TNF-α (tumour necrosis factor-α, Tnf), play a particularly important role in the inflammatory response. The transcript for IL-1b, an interleukin secreted by immune cells including monocytes and macrophages (Lopez-Castejon and Brough, 2011), was significantly upregulated in the AV node in HF ( Figures 7C and S20A). IL-1b is well known to be upregulated in HF and is thought to be involved in HF development (Van Tassell et al., 2015). The upregulation of Il1b in the AV node is therefore an important (and novel) finding. Figure  S20A shows changes in various transcripts involved with interleukins. TNF-α has again been implicated in HF development; for example TNF-α overexpression results in ventricular hypertrophy and dilatation, interstitial fibrosis, apoptosis, and a diminished ejection fraction (Hori and Yamaguchi, 2013). Although there was no change in Tnf in the AV node in HF, there were changes in other transcripts involved with TNF-α ( Figure S20B). For example, there was an upregulation of Tnfrsf12a (TWEAK), a positive regulator of cardiomyocyte proliferation (Novoyatleva et al., 2009), and Tnfrsf1a and Tnfrsf1b, which encode two receptors for TNF-α ( Figure S20B). Members of the transforming growth factor-β (TGF-β) superfamily are known to be activated in HF (Hanna and Frangogiannis, 2019). The TGF-β system stimulates myocyte hypertrophy and cardiac fibrosis (Kapur, 2011). In the AV node in HF there were significant changes in TGF-β superfamily transcripts: there was an upregulation of Tgfb1, Tgfb2, Gdf6 and Gdf15, but a downregulation of Gdf7, Gdf10, Bmp1, Bmp3 and Bmp5 ( Figure 7C).
Galectin-3 (Lgals3), a β-galactoside-binding lectin, has been proposed to have multifaceted functions in various pathophysiological conditions and can exert cytokine-like regulatory actions in immune cells (Jeon et al., 2010). Elevated serum galectin-3 levels have been detected in patients with almost all types of cardiovascular disease (including HF) and are considered as a biomarker of fibrosis and inflammation and may predict morbidity and mortality (Dong et al., 2018). Lgals3 was upregulated over three-fold in the AV node in HF ( Figure S20C). Galectin-3 exerts cytokine-like regulatory actions through the JAK-STAT pathway (Jeon et al., 2010) and the most abundant STAT transcript, Stat3, was significantly upregulated ( Figure S20C).

Remodelling of receptor and G protein transcripts in HF
Although Chrm2 (responsible for the M2 acetylcholine receptor) was unaffected in HF ( Figure  S22A), transcripts for one of the principal effectors of the M2 receptor, the ACh-activated K + channel, Kcnj3 and Kcnj5, were downregulated as already shown ( Figure 3C); there were also some changes in the G protein subunits mediating the effects ( Figure S22B). The adenosine A1 receptor also works via the ACh-activated K + channel; Adora1 was reduced by 22.2%, but only approached significance (P=0.072) ( Figure S22A). Unexpectedly, Adrb1, Adrb2, and Adrb3 (responsible for b-adrenergic receptors) were unaffected in HF ( Figure S22A), but as discussed above transcripts for some regulatory subunits of protein kinase A working downstream of the b-adrenergic receptors were affected. Adra1b, the most abundant a-adrenergic receptor transcript, was significantly downregulated as was Adra1d ( Figure S22A); interestingly Adra1b was more abundant than the badrenergic receptor transcripts ( Figure S22A). a-adrenergic receptors are believed to be cardioprotective in HF (Jensen et al., 2014;O'Connell et al., 2013). Transcripts for the mineralocorticoid receptor (Nr3c2; activated by aldosterone, deoxycorticosterone and cortisol) and thyroid hormone receptor-a (Thra) were downregulated in HF ( Figure S23A). Recently, sinus node cells have been shown to express functional glutamate receptors (Liang et al., 2021) and glutamate receptor transcripts were expressed in the AV node and some were downregulated in HF ( Figure  S23B).

Remodelling of sarcomeric transcripts in HF
There were significant changes of sarcomeric transcripts in the AV node in HF ( Figure S24). Just one change will be highlighted: in HF the ratio of Myh6:Myh7 (a-myosin heavy chain:b-myosin heavy chain) changed from 50.7 to 4.8.

Changes in transcription, translation, and mRNA transcript and protein breakdown underlying the remodelling of the AV node in HF
The remodelling of the AV node in heart failure will involve an increase in transcription and translation of some genes as well as the degradation of some mRNAs and proteins. Figure S25 shows a schematic diagram of the processes involved in in transcription, translation, and mRNA transcript and protein breakdown as well transcripts involved in these processes showing significant changes in HF. Transcription is controlled by histone acetyltransferases (HATs), which make chromatin accessible for transcription, and deacetylation of histones by histone deacetylases (HDACs) results in closed chromatin structure and the inhibition of gene transcription (although HDACs deacetylate many nonhistone proteins as well; McKinsey, 2011). Although there were no changes in HAT transcripts, there was a downregulation of three HDAC transcripts including Hdac5 in the AV node in HF ( Figure S25). Reduced expression of HDACs results in a more open chromatin state to allow gene transcription to occur. In the mouse, knockout of Hdac5 results in exaggerated cardiac hypertrophy in response to pressure overload and spontaneous pathologic hypertrophy with advancing age (McKinsey, 2011). However, HDACs also have a protective effect and HDAC inhibitors have been identified as a promising therapeutic approach in the treatment of HF (McKinsey, 2011). Eukaryotic initiation factors (EIFs) are involved in the initiation of translation and driving protein synthesis. In HF in the AV node there was an increase in numerous EIFs ( Figure S25) suggesting an increase in protein translation rates, a hallmark of HF (Simpson et al., 2020). RNA polymerases are essential for protein translation. In HF in the AV node, there was an increase in Polr1e, a subunit of polymerase I ( Figure S25). Polymerase I is involved in the synthesis of ribosomal RNA which is essential for protein synthesis (Goodfellow and Zomerdijk, 2013). There was also an increase in Mpp6 ( Figure S25), which is again involved in the generation of ribosomal RNA (Schilders et al., 2005). At steady-state, protein degradation has to match translation and, if there is an increase in protein translation rates in HF in the AV node, it is arguable that there should be a concurrent increase in protein degradation. Proteins are tagged for degradation by ubiquitination catalysed by ubiquitin ligases. Once ubiquinated, the protein is degraded by the proteasome. In HF in the AV node, there was an increase in numerous proteasome subunits ( Figure S25).

SUPPLEMENTARY DISCUSSION
In this study, an omics technology, RNAseq, was used to investigate changes in the whole transcriptome. In comparison, conventional hypothesis-driven-research is more focussed. Conventional hypothesis-driven-research channels the investigator's efforts and provides a framework for progressing knowledge in a methodical stepwise manner. However, it also blinkers the investigator to the complexity of biological systems potentially leading to erroneous conclusions (Boyett and Lundby, 2020). Omics technologies allows the alternative research strategy of datadriven-discovery, and does not constrain the view of the investigator (Boyett and Lundby, 2020).

Comparison of changes in ion channel expression in the AV node in this and earlier studies of HF
In the mouse model of hypertension-induced HF in this study, there was AV node dysfunction as indicated by a prolongation of the PR interval ( Figure 1G). AV node dysfunction has been observed in other animal models by us: (i) in a rabbit model of pressure and volume overloadinduced HF, there is an increase in the PR interval, (ii) in a rat model of pulmonary hypertension and right-sided HF there is an increase in the AH interval, Wenckebach cycle length and AV node effective and functional refractory periods measured in the isolated Langendorff perfused heart and a 50% incidence of AV block in the isolated AV node, (iii) in a rat model of ageing there are increases in the AH interval, Wenckebach cycle length and AV node effective refractory period in the isolated AV node, (iv) in a horse model of exercise training there is an increase in the PR interval and incidence of second degree AV block, and (v) in a mouse model of exercise training there are increases in the PR interval, Wenckebach cycle length, and AV node effective refractory period (Nikolaidou et al., 2015;Temple et al., 2016;Saeed et al., 2018;Mesirca et al., 2021). In the mouse model of pressure overload-induced HF in this study, the AV node dysfunction most likely was the result of ion channel remodelling in the AV node ( Figure 3) and this is also true of the other animal models. In the rabbit model of pressure and volume overload-induced HF, there is a downregulation of Hcn1, Cacna1d (Cav1.3), Gja5 (Cx40) and Gja1 (Cx43) transcripts (Nikolaidou et al., 2015). In the rat model of pulmonary hypertension and right-sided HF, there is a widespread downregulation of ion channel and related genes, e.g. Hcn1, Hcn2, Hcn4, Cacna1(Cav1.2) and Cacna1d (Cav1.3) (Temple et al., 2016). In the rat model of ageing, there is a downregulation of HCN4, Nav1.5, RYR2 and Cx43 proteins (Saeed et al., 2018). In the horse model of exercise training there is a downregulation of HCN4 and Cav1.2 protein (Mesirca et al., 2021). In the mouse model of exercise training there is a downregulation of various ion channels transcripts: Hcn2, Hcn4, Cacna1(Cav1.2), Cacna1g (Cav3.1), Cacna1h (Cav3.2), Ryr2, and various K + channel and connexin transcripts; there is also a downregulation of HCN4 and Cav1.2 protein and ICa,L and If (Mesirca et al., 2021). It is concluded from the present study as well as the earlier studies that AV node dysfunction in HF and other conditions is likely to be the result of a transcriptional downregulation of key ion channels involved in AV node function.

Limitation of the study
In this study, the C57Bl6/N mouse TAC model is described as a model of HF as we did previously (Yanni et al., 2020). Previous studies have supported the use of TAC in C57Bl6/N mice as a model of HF (Zi et al., 2019). In this study, HF was defined based on a multitude of symptoms: reduced function (reduced ejection fraction, reduced fractional shortening), increased heart size (increased heart weight, heart weight/body weight ratio, left ventricular mass, left ventricular end diastolic diameter, and left ventricular end systolic diameter), reduced body weight, reduced heart rate, and outward signs of discomfort such as laboured breathing and lack of movement. These signs were evident in most if not all the TAC mice in this study (Figures 1 and S1). Neverthless, the outcomes of TAC were heterogeneous as shown in Figures 1 and S1. For example, the ejection fraction of all the TAC mice was lower than that of the control mice, but it still varied from 29.1 to 61.5%. The criterion for HF with preserved ejection fraction is generally an ejection fraction >50% (Pfeffer et al., 2019), whereas the criterion for HF with reduced ejection fraction is <40% (Murphy et al., 2020). Five of the 10 mice had an ejection fraction >40% and of these three had an ejection fraction >50%. It is possible that some mice displayed HF with a preserved ejection fraction. Hypertrophy cardiomyopathy is characterised by cardiac hypertrophy, a non-dilated left ventricle and a normal or increased ejection fraction, which is not evident in our study (Marian and Braunwald, 2017).
In this study, transcripts were measured in tissue taken from the Triangle of Koch lying between the coronary sinus, the tendon of Todaro and the tricuspid valve annulus (Li et al., 2008). This will not only include the compact node made up of N (nodal) cells, but also transitional tissue made up of AN (atrio-nodal) cells, the inferior nodal extension made up of N cells, and perhaps the start of the penetrating bundle made up of NH (nodal-His) cells (Inada et al., 2009). Whereas the compact node and inferior nodal extension are made up of typical nodal cells, AN and NH cells are more transitional in nature. In a previous study of the rat, we used laser-assisted microdissection to collect tissue from each of the regions and quantitative PCR (qPCR) to measure the expression of selected transcripts (Temple et al., 2016). This, however, is labour intensive and time consuming and would also be more difficult in a smaller animal (the mouse rather than the rat), which is why in this study expression of transcripts was measured in Triangle of Koch biopsies as we did in a previous study of the mouse (Mesirca et al., 2021). The limitation of this is that the measured transcript expression will be determined by the expression within the different cell types in the Triangle of Koch. To estimate what is determining the measured transcript expression in the present study, five marker transcripts were compared in this study (of the Triangle of Koch) and our previous study (of the various tissues making up the AV junction). The marker transcripts are Hcn1 and Hcn4 (known to be highly expressed in the AV node and poorly expressed in the atrial muscle) and Scn5a, Kcnj2 and Gja1 (known to be poorly expressed in the AV node and highly expressed in the atrial muscle). The table below shows the ratio of the nodal markers (Hcn1 or Hcn4) to the atrial muscle markers (Scn5a, Kcnj2 and Gja1) in the present study and our previous study (Temple et al., 2016) of the various tissues of the AV junction. The expression of the marker transcripts in the present study is unlike that of atrial muscle, but only in one instant was it like that of the compact node (Hcn1/Scn5a); in general, the expression of the marker transcripts in the present study is consistent with the expression in N and AN cells. A limitation of this analysis is that mouse tissue is being compared to rat tissue. The various signalling pathways discussed above are known to be present in various cell types including immune cells as well as in cardiomyocytes. The Hippo pathway in various cell types is known to be involved in activation of the immune system following cardiac injury (Mia and Singh, 2019). WNT signalling in immune cells plays an important role in immune cell regulation (Chae and Bothwell, 2018). CaMKII is a mediator of inflammatory processes in the heart: CaMKII signalling in the immune system is responsible for the pro-inflammatory cytokine production in macrophages, and CaMKII signalling in the heart influences the degree of the inflammatory response (Beckendorf et al., 2018). The p38-MAPKα/β pathway is involved in the inflammatory response (Clerk and Sugden, 2006) and p38 MAPK overexpression in the heart has been shown to induce gene expression resulting in myocardial cell proliferation, inflammation, and fibrosis (Tenhunen et al., 2006). Therefore, the changes in the signalling pathway transcripts observed in this study may be occurring in for example immune cells rather than cardiomyocytes -further study is required to clarify this.

A final comment
Although hypothesis-driven research has many advantages, it blinkers the investigator to the complexity of biological systems potentially leading to erroneous conclusions, whereas omics technologies and data-driven discovery does not constrain the view of the investigator in this way (Boyett and Lundby, 2020). The current study is a good example of this: it has revealed that HF likely causes AV node dysfunction not as a result of a change in a single or small number of molecules. Instead, the AV node dysfunction likely involves widespread changes in many different cellular systems. This has implications for the development of new treatments.

ADDITIONAL SUPPLEMENTARY FILES All transcripts.xlsx
For all transcripts detected, this file lists: gene name; mean, SEM and n for both control and HF; ratio of the HF mean to control mean expressed as a percentage; P value; Benjamini-Hochbergadjusted P value.
TGF-β and WNT signaling pathways in cardiac fibrosis: non-coding RNAs come into focus.

S l c 2 7 a 5 ( F A T P 5 )
S l c 2 7 a 6 ( F A T P 6 )      Heart failure/control (%)

n 1 P x d n M a t n 4 T h b s 2 I g s f 1 0 I g f b p 5 L g i 4 T n x b D p t C h a d l V i t K c p L a m b 3 S n e d 1 I g f b p l 1 C o c h N t n 3 S b s p o n E f e m p 1 C i l p 2 V w a 3 b L g i 3 L g i
Interleukin pathway

Crlf1
Nppb R 2 =0.92 P=0.0023 Figure S21. Relationship between two HF markers (Nppa and Nppb) and a macrophage marker (Cd14), a cytokine transcript (Crlf1), and a transcript for a cytokine-like molecule known to be associated with HF (Lgals3). Each point corresponds to one of six pooled samples (three control and three HF samples). Each pooled sample is made up of AV node biopsies from three mice. The data have been fitted with a straight line and the R 2 value and the P value (of a slope of zero) are shown. In each case the control samples are clustered near the origin.  Heart failure/control (%) Figure S25. Transcription, translation, and mRNA transcript and protein breakdown. Top, schematic diagram of the cycle of transcription, translation, and RNA and protein degradation (modified from Wang et al., 2021b). Bottom, expression of transcripts involved in transcription, translation, and mRNA transcript and protein breakdown in HF (as a percentage of that in control). Colours correspond to the different parts of the pathway above. Red dotted line corresponds to 100%.