New aspects of endocrine control of atrial fibrillation and possibilities for clinical translation

Abstract Hormones are potent endo-, para-, and autocrine endogenous regulators of the function of multiple organs, including the heart. Endocrine dysfunction promotes a number of cardiovascular diseases, including atrial fibrillation (AF). While the heart is a target for endocrine regulation, it is also an active endocrine organ itself, secreting a number of important bioactive hormones that convey significant endocrine effects, but also through para-/autocrine actions, actively participate in cardiac self-regulation. The hormones regulating heart-function work in concert to support myocardial performance. AF is a serious clinical problem associated with increased morbidity and mortality, mainly due to stroke and heart failure. Current therapies for AF remain inadequate. AF is characterized by altered atrial function and structure, including electrical and profibrotic remodelling in the atria and ventricles, which facilitates AF progression and hampers its treatment. Although features of this remodelling are well-established and its mechanisms are partly understood, important pathways pertinent to AF arrhythmogenesis are still unidentified. The discovery of these missing pathways has the potential to lead to therapeutic breakthroughs. Endocrine dysfunction is well-recognized to lead to AF. In this review, we discuss endocrine and cardiocrine signalling systems that directly, or as a consequence of an underlying cardiac pathology, contribute to AF pathogenesis. More specifically, we consider the roles of products from the hypothalamic-pituitary axis, the adrenal glands, adipose tissue, the renin–angiotensin system, atrial cardiomyocytes, and the thyroid gland in controlling atrial electrical and structural properties. The influence of endocrine/paracrine dysfunction on AF risk and mechanisms is evaluated and discussed. We focus on the most recent findings and reflect on the potential of translating them into clinical application.

by a combination of direct electrophysiological effects and indirectly through AF-associated risk factors, the effects of cortisol deficiency on atrial electrophysiology and AF are unknown. Recent studies have identified the importance of GR-transcriptionally induced glucocorticoidinduced leucine zipper (GILZ or tsc22d3) protein. GILZ mediates multiple effects of glucocorticoids that has not been linked to AF potentially due to more selective, as opposed to glucocorticoids, effects. 13 GH is secreted by the anterior pituitary and has direct as well as indirect anabolic and positive inotropic effects mediated via the GHdirected hepatic secretion of insulin-like growth factor (IGF)-1. 14 Both GH excess and deficiency have been associated with increased cardiac arrhythmias, including AF, cardiovascular morbidity, and mortality. 15 Chronic GH excess, most often caused by a pituitary GH-secreting adenoma, leads to acromegaly, with typical morphological and clinical features. Elevated GH is associated with left ventricular (LV) hypertrophy, left-sided valvular heart disease, and other cardiovascular risk factors for AF, including diabetes, coronary artery disease, hypertension, and dyslipidaemia. 16 The mechanism underlying acromegaly-associated AF has not been fully elucidated but likely involves left atrial (LA) enlargement with pro-fibrillatory structural remodelling ( Figure 2).
GH deficiency, a common manifestation of hypopituitarism, presents with growth retardation in children and as a cardiometabolic syndrome in adults. 15 Patients with GH deficiency GH develop hypertension, reduced LV mass (Figure 2), 17 and LA structural remodelling that may mediate increased AF risk in patients with GH deficiency. Cardiomyocytes express GH and IGF-1 receptors but there are no reported direct effects of GH/IGF-1 on atrial electrophysiology. 18 FSH and LH stimulate oestrogen production in women (ovaries) and testosterone in men (testis), among other functions. Changes in cardiac electrophysiology across the menstrual cycle have been described, with oestrogen having APD-prolonging effects (follicular phase) while progesterone tends to shorten APD (luteal phase) by decreasing and increasing K þ repolarizing currents, respectively 19 ; however, oestrogen abnormalities have not been formally linked to a higher risk of AF.
Low testosterone levels can result from primary (testis), secondary (pituitary), or tertiary (hypothalamus) male hypogonadism in young individuals, or as the result of normal ageing in older men. 20 Males with lower testosterone levels have worse cardiovascular outcomes in general, including higher rates of AF. 21 Low testosterone levels are associated with several AF-related cardiovascular risk factors such as diabetes, premature coronary artery disease, dyslipidaemia, and obesity. Accordingly, recent work suggests that transgender women on hormonal therapy might be at increased risk of AF 22 ; however, the available data are scarce, and a larger database is needed to confirm this association.
Furthermore, proinflammatory markers like C-reactive protein (CRP) and interleukin-6 (IL-6) have an inverse correlation with testosterone levels, 23 suggesting that inflammation may contribute to pro-fibrillatory atrial structural remodelling. Testosterone affects cardiac electrophysiology, as castrated rats showed increased expression of the ryanodine receptor type-2 (RyR2), sodium (Na þ )/Ca 2þ exchanger (NCX), late Na þ current (I NaL ), APD prolongation, and higher AF susceptibility vs. control animals. 24,25 Androgen receptor knock-out (KO) rats have reduced resting membrane potential, increased APD, and enhanced sensitivity to isoproterenol-induced delayed afterdepolarizations, 26 that are partly reversed by testosterone replacement therapy. 20 A TSH-producing adenoma is the most common cause of central hyperthyroidism, a rare cause of thyrotoxicosis. Similarly, central hypothyroidism is much less common than its primary counterpart. Although poorly characterized, the cardiovascular consequences of central hypo-/ hyperthyroidism are expected to parallel those observed with primary hypo-/hyperthyroidism, as described below.
Prolactin stimulates milk production in the gravid woman and has not been association with AF.

The adrenals and AF
The adrenal, or suprarenal, glands are composed of two embryologically distinct and functionally independent cortical and medullary units. The adrenal cortex contains three layers secreting aldosterone (zona glomerulosa), cortisol (zona fasciculata), and androgens (zona reticularis).
The adrenal medulla is part of the sympathetic nervous system and releases epinephrine and norepinephrine.
Aldosterone is normally secreted in response to hypovolemia and hyperkalaemia and binds mineralocorticoid receptors (MRs) to regulate Na þ /K þ /H þ homeostasis by activating the epithelial Na þ channel (ENaC) in the nephron distal tubule and collecting duct. 27 MRs are also expressed on other cell types, including atrial and ventricular cardiomyocytes. The association between aldosterone and AF has most clearly been demonstrated in patients with primary hyperaldosteronism (PA; renin-independent aldosterone production), the most common cause of secondary hypertension. 28 AF is more prevalent in patients with PA (7.3%) vs. matched patients with essential hypertension (0.6%) 29 , adrenalectomy lowers the PA-associated risk of AF 30 and the MR-specific antagonist eplerenone is associated with a 42% relative risk reduction of AF, 31 all suggesting a potential role of aldosterone in the pathogenesis of AF independent of upstream regulatory hormones (i.e. renin/angiotensin-II).
The mechanisms linking hyperaldosteronism and AF are complex and incompletely understood ( Figure 3). However, chronic excess aldosterone leads to LA enlargement and remodelling, indirectly via hypertension and diastolic dysfunction and/or directly via blood pressure-independent effects. [32][33][34] Aldosterone-binding to cardiac macrophage MRs increases expression of the profibrotic markers transforming growth factor-beta 1 (TGF-b1), matrix metalloproteinase-12, tumour necrosis factor-alpha (TNFa), and plasminogen activator inhibitor-1 (PAI-1) 35,36 , promotes macrophage-mediated oxidative stress 37 and stimulates the Keep1/Nrf2-dependent cardiac fibroblast to myofibroblast transformation 38 contributing to pro-fibrillatory LA structural remodelling. Conduction time and P-wave duration are increased in a rat model of hyperaldosteronism, compatible with increased atrial fibrosis and slowed conduction. 39 Aldosterone also has direct electrophysiological effects, the clinical significance of which remains to be fully elucidated. Aldosterone administration to rats caused APD shortening, mediated by an increase in Kir2.1 (inward rectifier K þ current, I K1 ) and Kv1.5 (ultrarapid delayed rectifier K þ current, I Kur ) expression; these changes were reversed by the MR antagonist spironolactone. 40 Conversely, ventricular APD was prolonged in a MR-overexpression model because of the downregulation of transient outward K þ current (I to ) and upregulation of L-type Ca 2þ current (I CaL ). 41 It has been shown that ventricular I CaL magnitude correlates with aldosterone levels 42,43 and MR-activation increases sarcoplasmicreticulum (SR) Ca 2þ -sparks and delayed afterdepolarizations, 44 Figure 3 Hyperaldosteronism and AF. (A) Hypovolemia and hyperkalaemia are the primary physiological stimuli for adrenal aldosterone secretion, which acts on the nephron distal tubule and collecting duct to retain Na þ and excrete K þ . (B) Mechanism of aldosterone-related AF. Hyperaldosteronism causes angiotensin-independent hypertension and left atrial (LA) inflammation, leading to pro-fibrillatory LA remodelling. It also produces pro-AF electrical remodelling in the form of LA action potential-shortening, increased sarcoplasmic reticulum Ca 2þ sparks, and delayed afterdepolarizations. Sustained AF may potentiate the effects of hyperaldosteronism by upregulating of the mineralocorticoid receptor (MR) on AF atrial cardiomyocytes (CMs). Furthermore, AF increases 11b-hydroxysteroid dehydrogenase type 2 (11b-HSD2), which metabolizes cortisol, thereby increasing MR occupancy by aldosterone. implicated in AF pathophysiology. A mouse model of spontaneous AF was associated with increased 11b-hydroxysteroid dehydrogenase type 2 (11b-HSD2), which inactivates cortisol, thereby allowing for increased MR occupancy by aldosterone. 45 Patients with AF have upregulated expression of MRs vs. sinus-rhythm controls 46 and MR antagonists reduce the risk of AF in heart failure patients. 31 Hence, AF may itself potentiate aldosterone's proarrhythmic effects.
Pheochromocytomas are epinephrine and norepinephrine-secreting adrenal tumours. The excess catecholaminergic state leads to hypertension, myocardial ischaemia/or, and cardiomyopathy. 47 Cardiac arrhythmias are common in pheochromocytoma and AF was documented in 11.3% of patients with this condition, all of which responded to tumour resection. 48

Obesity and AF
Obesity and AF are both reaching epidemic levels. Obesity can result from endocrine abnormalities, and adipose tissue secretes a number of bioactive hormones. 49 Obesity is a well-established AF risk factor. 45 Weight loss reduces the risk of AF in obese patients improves outcomes after AF catheter ablation and reverses obesity-related electrical remodelling. 50 The relationship between obesity and AF has been framed into a two-component model: the corporal load model and the lipotoxicity model ( Figure 4). 51 The corporal load model stipulates that the increase in hemodynamic load imposed by excess weight leads to left ventricular (LV) hypertrophy, diastolic dysfunction with secondary LA enlargement, and pro-fibrillating remodelling. Interestingly, excess lean body mass, not merely adipose mass, is an important mediator of obesity-associated AF risk. 52 The lipotoxicity model refers to the direct proinflammatory and profibrotic states associated with obesity. Obesity induces pro-inflammatory signalling in the atria 53 that can promote both ectopic firing and an AFmaintaining electrical and structural substrate, ultimately leading to AF. 54 Visceral adiposity is associated with increased blood leukocyte count, CRP, IL-6, and TNF-a. 55,56 Accelerated fibrogenesis mediated by the TGF-b1, connective tissue growth factor (CTGF), and endothelin-1 systems, 57 among others, contributes to the proarrhythmic lipotoxic effect. Epicardial adipose tissue (EpAT) is in direct contact with the epicardium and shares its blood supply, positioning it to affect atrial electrophysiology through paracrine and vasocrine interactions. EpAT volume correlates with areas of atrial fibrosis, slow conduction, electrogram fractionation, and lateralization of connexin (Cx)-40, 58 suggesting a direct effect on atrial electrophysiology. Similar to visceral fat, EpAT secretes metabolically active (e.g. free fatty acids), angiogenic (e.g. vascular endothelial growth factor), growth (e.g. activin A), and remodelling (TGF-b1/2 and MMPs) factors, as well as inflammatory cyto-and chemokines (e.g. IL-6 and PAI-1), and adipokines (e.g. leptin). 59 While leptin is implicated in angiotensin-II-mediated pro-fibrillatory atrial remodelling, 60,61 another adipokine, resistin, correlates with clinical AF risk. 62 Adipocytes also secrete neprilysin, a neutral endopeptidase that degrades cardioprotective endogenous natriuretic peptides (NPs), 63 that positively correlates with body mass index (BMI) 64 and regulates angiotensin-II concentrations in human adipose tissue. 65 An overproduction of aldosterone, 66 linked to AF pathogenesis, shorter atrial, and pulmonary vein refractory periods, conduction slowing, and heterogeneity 57,67 are also found obesity and may be involved in obesityassociated AF.

Renin-angiotensin system and AF
Hypertension, an important risk factor for AF, [68][69][70][71] is often associated with activation of the renin-angiotensin system (RAS). 72 The RAS system is a neuroendocrine axis involving kidney production of renin that converts liver-produced angiotensin into angiotensin-I, which is subsequently converted into active circulating angiotensin-II in the lungs. Hypertension leads to atrial remodelling as indicated by LA enlargement and prolongation of P-wave duration. 73,74 Angiotensin-II infusion produces a rapid increase in systolic blood pressure (exceeding 140-150 mmHg) [75][76][77] and a substantial increase in susceptibility to AF in mice. [75][76][77][78][79] Enhanced AF-susceptibility in angiotensin-II infused mice occurs in association with atrial enlargement, atrial fibrosis, and prolonged P-wave duration. 75,80,81 Consistent with Pwave prolongation in vivo, optical mapping demonstrates conduction slowing in the right atrium (RA) and LA of angiotensin-II infused mice. 75 RA and LA APD are prolonged and sinoatrial node function is impaired in angiotensin-II infused mice ( Figure 5A). 75 Notably, atrial tachyarrhythmia itself induces angiotensin-II type 1 receptor-mediated oxidative stress (mainly due to increased nicotinamide adenine dinucleotide phosphate oxidase activity, LOX-1 upregulation, and F 2 -isoprostane generation) in the ventricular myocardium, negatively impacting on its function. 82 Ion-channel remodelling may explain the electrophysiological changes associated with AF promotion by angiotensin-II. LA I Na is reduced by approximately 50% in angiotensin-II infused mice, apparently via enhanced PKCa activity as dialysis with BIM1 (a PKC inhibitor) normalized I Na density and activation kinetics. 69,71 APD-prolongation occurred in conjunction with decreased outward K þ -current (I K ), attributed to reductions in I to and I Kur independently of a change in K v 4.2, K V 4.3, KChIP2, and K v 1.5 protein levels.
AF following angiotensin-II infusion in mice is also associated with oxidative stress, leading to oxidation of Ca 2þ -calmodulin-dependent protein kinase II (CaMKII). 78 CaMKII-oxidation leads to pathological, constitutively active CaMKII-signalling. 83 Oxidized CaMKII expression is increased in both AF patients and mice infused with angiotensin-II. 78 CaMKII oxidation causes arrhythmogenic alterations in SR Ca 2þ -handling, with increased Ca 2þ -sparks leading to delayed afterdepolarizations. Knock-in mice resistant to CaMKIId oxidation are protected from Ca 2þ -mishandling and show less AF inducibility. 78 Angiotensin-II infusion also causes atrial interstitial fibrosis 75,77 ( Figure 5A), resulting from altered extracellular matrix (ECM) remodelling by MMPs and TIMPs under the influence of oxidative stress and inflammation. 78,79

Natriuretic peptides in AF
Natriuretic peptides (NPs) are cardioprotective hormones that play important roles in regulating cardiac electrophysiology and arrhythmogenesis. 84,85 NPs modulate atrial AP morphology and alter atrial conduction patterns by regulating ion-channel function. [84][85][86][87][88] NPs elicit their effects by binding to NP receptors (NPRs), including NPR-A, NPR-B, and NPR-C. 89 NPR-A and NPR-B are guanylyl cyclase-linked receptors that modulate cGMP signalling, while NPR-C is coupled to the inhibitory G protein (Gi) and phospholipase C signalling. 84,90,91 NPR-C is highly expressed in the atria, 86,92 and recent studies have identified an essential role for NPR-C in regulating atrial conduction and AF inducibility. 77,92 NPR-C knockout (NPR-C -/-) mice display increased susceptibility to burst pacing-induced AF in association with impaired New aspects of endocrine control of atrial fibrillation and possibilities for clinical translation atrial conduction. Atrial interstitial fibrous-tissue content is increased in NPR-C -/mice, whereas no differences in atrial AP morphology occur. 92 In contrast, no ventricular arrhythmias or ventricular fibrosis were observed in NPR-C -/mice, 92 further indicating that NPR-C is particularly important for regulation of atrial structure and function.
NPR-C plays a modulating role in angiotensin-II mediated AF. 77 Angiotensin-II infusion in NPR-C -/mice produces enhanced effects on AF vulnerability and duration, P-wave duration, and atrial conduction. Reductions in AP upstroke velocity (V max ) and I Na , as well as APD prolongation, are seen, particularly in LA cardiomyocytes. Angiotensin-II infusion also produced larger increases in PKCa protein expression in NPR-C -/mice, as well as enhanced RA and LA fibrosis. 77 Co-treatment of wild-type mice with angiotensin-II and cANF (a synthetic, selective NPR-C agonist 90 ) reduces AF burden, while improving atrial conduction, attenuating atrial fibrosis, and improving AP properties. 77 These findings are consistent with other work showing that NPR-C plays protective roles in the cardiovascular system. 76,[93][94][95] On the other hand, one study using transverse aortic constriction and TGF-b-overexpression found that the absence of NPR-C was protective against AF and atrial fibrosis. 96 The basis for these contradictory observations is unclear and further studies are warranted.
Studies showed that mutations in the atrial natriuretic peptide (ANP) gene are linked to AF. 97,98 A family with familial AF was shown to have a frameshift mutation in the NPPA gene (encoding ANP) that results in a mutated ANP (mANP) circulating at concentrations 5-10 times greater than wild-type ANP because of increased resistance to proteolytic degradation. 98,99 While ANP has been shown to increase funny current (I f ) in human atrial cardiomyocytes and predispose to AF, 100 ANP and mANP have also been found to demonstrate opposite effects on mouse and human atrial cardiomyocytes. 101 Specifically, ANP increased V max , APD, and I Ca,L in isolated atrial cardiomyocytes via the NPR-A receptor. In intact mouse atrial preparations, ANP speeded atrial conduction and increased atrial effective refractory period (AERP). In contrast, mANP decreased atrial V max , shortened atrial APD, decreased atrial I Ca,L , slowed atrial conduction, and shortened AERP. These effects were mediated by the NPR-C receptor, as the effects of mANP were absent in NPR-C -/mice. ANP and mANP also had opposing effects on I Ca,L in human RA cardiomyocytes. Finally, mANP administration caused re-entrant conduction patterns, ectopic firing, and disorganized conduction in mouse atria exposed to programmed stimulation, an effect not seen with ANP. These studies suggest that mANP is proarrhythmic in association with a shift in the balance between NPR-A and NPR-C mediated NP-signalling. Mice expressing the same nppa frameshift mutation show increased AF burden in association with APD shortening, in association with ion-channel remodelling, including changes in amplitude and expression of Na þ , Ca 2þ , and K þ channels. 94 Collectively, the available data indicate that this frameshift nppa mutation increases Angiotensin-II also causes right and LA fibrosis. These alterations lead to conduction abnormalities and increased susceptibility to AF. Loss of NPR-C leads to worsened ion channel remodelling and atrial fibrosis, as well as enhanced AF susceptibility, while NPR-C activation prevents some ion channel remodelling, reduces right and LA fibrosis, and decreases AF burden. (B) T1DM (Akita mice) is associated with reductions in AP V max due to reduction in atrial I Na as well as increases in AP duration due to reduction in I Kur . T2DM (db/db mice) show increases in AP duration due to reduction in both I to and I Kur while I Na amplitude and AP V max are not altered. Both T1DM and T2DM are associated with increased atrial fibrosis. These alterations lead to conduction abnormalities and increased susceptibility to AF. Insulin treatment in T1DM prevents reductions in atrial I Na and reduces atrial fibrosis leading to improved conduction and reduced AF susceptibility.
New aspects of endocrine control of atrial fibrillation and possibilities for clinical translation susceptibility to AF in association with increased circulating mANP levels and shortening of the atrial AP, which could decrease the wavelength for re-entry ( Figure 5A).

AF in diabetes mellitus
Type-1 (T1DM) and type-2 (T2DM) diabetes mellitus (DM) are metabolic disorders associated with hyperglycaemia and changes in insulin production and signalling.102-104 T1DM is characterized by the loss of insulin-producing b-islet cells in the pancreas and deficient insulin generation. T2DM, which often occurs in association with obesity, is characterized by insulin resistance in peripheral tissues while insulin can still be produced in the pancreas. 102 T2DM can ultimately lead to insulin insufficiency requiring insulin therapy.
AF is prevalent in both T1DM and T2DM. 69,71,105 T1DM is associated with atrial electrical and structural remodelling. 69 Experimental studies have evaluated atrial effects in genetic (Akita mice) or chemically induced [streptozotocin (STZ) or alloxan] animal models, of T1DM 106 that are characterized by substantial increases in AF susceptibility and duration. 107,108 Akita and STZ mice had increased P-wave duration; Akita mice also had impaired RA and LA conduction, 107 reduced atrial V max , and prolonged APD. These AP morphology changes occurred in association with reductions in I Kur and I Na , associated with decreased SCN5a gene and Na V 1.5 protein expression, as well as a loss of phosphoinositide 3-kinase (PI3K) signalling via the second messenger PIP 3 . Strikingly, insulin treatment protected against changes in I Na , but not the changes in I Kur , in Akita mice. Chronic insulin treatment increased Na V 1.5 protein levels, atrial I Na density, and AP upstroke velocity. Insulin could also increase atrial I Na and AP upstroke velocity acutely via the rapid activation of PIP 3 signalling. These effects of insulin on atrial I Na were associated with increases in atrial conduction velocity and were sufficient to reduce the AF burden. Consistent with previous work on the PI3K-and PIP 3mediated effects on Na þ -channel function, 109 these studies revealed a critical role for insulin in regulating atrial electrophysiology and AF susceptibility via effects on atrial Na þ channels in T1DM, though the basis for I Na dysregulation in T1DM remains unclear. Structural profibrotic remodelling in T1DM [110][111][112] is increased in the atria of Akita mice, 110 STZ-treated rodents, [111][112][113] and alloxan-treated rabbits, 114 that in STZtreated rats was prevented by inhibition of the type-1 angiotensin-II receptors, 112 suggesting a critical role of angiotensin-II in T1DM-related atrial fibrosis ( Figure 5B). STZ-treated rats showed reduced velocity and increased heterogeneity of the LA conduction associated with a reduced LA Cx40 expression, leading to increased arrhythmogenesis, 113 while atrial Cx40 or Cx43 mRNA expression were unaltered in Akita mice. 107 T2DM, which accounts for up to 90% of DM-patients, continues to increase at epidemic proportions in association with rising rates of obesity and metabolic syndrome. 115 AF is prevalent in T2DM. 105,116 Clinical studies in T2DM patients have shown that alterations in Ca 2þ -handling may contribute to atrial remodelling. 117 T2DM patients have atrial interstitial fibrosis and increased EpAT, potentially infiltrating atrial myocardium,118-123 that could lead to impaired electrical conduction and AF ( Figure 5B). While mechanistic studies in animal models are limited, recent work in T1DM and T2DM mouse models demonstrated that AF promotion is related to pro-arrhythmic activation of CaMKII (due to oxidative stress-mediated oxidation and O-GlcNAcylation of CaMKII) 124 that would potentially affect multiple ion currents. 83 AF in T2DM was recently investigated in db/db mice, which carry a mutation in the leptin receptor leading to obesity, insulin resistance, and hyperglycaemia. 125 The db/db mice have increased susceptibility to burst-pacing induced AF, associated with increased P-wave duration and AERP, reduced RA and LA conduction velocity, and prolonged heterogeneous APD. These changes were accompanied by a decrease in I to (associated with reduced expression of Kcnd2 mRNA and K v 4.2 protein) and suppressed I Kur, occurring in the presence of unchanged K V 1.5 expression. Zucker diabetic fatty (ZDF) rats also showed increased AF susceptibility and APD, 126 associated with reduced atrial I to , I Kur , and I Ca,L currents, and respective channel protein subunits Kv4.3, Kv1.5, and Cav1.2, indicating some model-specific differences. In contrast to Akita (T1DM) mice, 107 atrial I Na was not reduced in db/db atrial myocytes. 125 The only alteration observed in atrial I Na in db/db mice was a shift in steady-state inactivation that resulted in a larger I Na window current, which could contribute to the prolongation of APD. This observation identifies potentially important differences in electrical remodelling between T1DM and T2DM that may have important implications for AF therapy in these related, but distinct conditions.
Animal models of T2DM also consistently display atrial structural remodelling, including fibrosis, lipidosis, and inflammation ( Figure 5B), 125,127-130 which has been shown to promote AF. 54 Adipokines (cytokines with pro-inflammatory properties) like leptin are also implicated in the atrial fibrosis of diabetic mice, 131 while cathepsin-A (a proteolytic enzyme active in the extracellular space) contributes to atrial fibrosis in ZDF rats. 127 Gene expression of Cx40 and Cx43 remained unchanged in both db/db mice and ZDF rats, 125,127 yet lateralization of Cx43 was observed in ZDF rats 127 that could underlie higher conduction heterogeneity.

Thyroid dysfunction in AF
Thyroid disease has a large number of well-characterized cardiac manifestations and both hypo-and hyperthyroidism have been associated with worse cardiovascular outcomes. 132 Although clinical (low TSH, elevated T 4 ) and subclinical (normal TSH, elevated T 4 ) hyperthyroidism, even with marginally increased T 4 levels, have been linked to a higher risk of AF, [133][134][135] overt hyperthyroidism is present in less than 1% of patients with new-onset AF. 136 Conversely, antiarrhythmic treatment with amiodarone is itself an important potential cause of hyperthyroidism in cardiac patients, so-called amiodarone-induced thyrotoxicosis. 137 Elevated thyroid hormone (TH) levels have also been associated with increased atrial premature depolarizations and supraventricular tachyarrhythmias. 138,139 Hypothyroidism appears to have relatively protective effects, especially against malignant ventricular arrhythmias, and a much less robust association with AF. 140 Thyrotropin-releasing hormone (TRH) release from the hypothalamus stimulates secretion of TSH by the anterior-pituitary and subsequent release of tetraiodothyronine (thyroxine; T 4 ) and in lesser amounts (1:9 ratio), triiodothyronine (T 3 ) by the epithelial cells of the thyroid gland ( Figure 6). The enzyme 5 0 -iodinase converts T 4 into T 3 , the more metabolically active TH. T 3 and T 4 inhibit TRH and TSH release, forming a negative-feedback loop. TH actions can be genomic or nongenomic ( Figure 6). Genomic effects are mediated as T 3 enters the nucleus and interacts with the nuclear thyroid a receptor-1 (TRa1), which binds the thyroid release elements (TREs), promoting/repressing transcription of TH-regulated genes like sarcoplasmic reticulum Ca 2þ adenosine triphosphate (SERCA2), phospholamban, Na þ /K þ ATPase, NCX, selected voltage-gated K þ currents, I CaL , and b 1 -adrenergic receptor. 141 Non-genomic effects (reviewed elsewhere 142 ) have a rapid onset of actions that are transcription/translation-independent and are mediated by extra-nuclear receptors, structurally related to the thyroid receptorlike integrin avb3, cytoskeleton, mitogen-activated protein kinase 1 =2 and PI3K.
Hyperthyroidism leads to so-called high-output heart failure characterized by hyperdynamic congestive failure in patients with normal baseline LV-function. 143 Left untreated, heart failure leads to LAenlargement, activation of the RAAS, and increased sympathetic tone ( Figure 6) 142 and hypertension. 142 THs also have important electrophysiological effects. Hyperthyroidism leads to sinus tachycardia from increased rates of diastolic depolarization in sinoatrial cells; hypothyroidism has the opposite effect. 144,145 Similarly, APD is consistently prolonged in animal models of hypothyroidism, [146][147][148][149][150] whereas hyperthyroidism shortens APD. 147,148,151 There are no documented effects of thyroid disease on the resting membrane potential. Beyond these well-characterized macroscopic electrophysiological changes, there is substantial heterogeneity in the reported TH-induced ion-channel remodelling.

Depolarizing currents
There are no documented effects of hypo-/hyperthyroidism on I Na , although one study reported increased/decreased conduction velocity in hyper-/hypothyroid rabbit atria, respectively. 146 I CaL has been found to be increased, 148,152,153 unchanged 154 or decreased 155,156 in hyperthyroid models. Interestingly, one study found hyperthyroidism to increase I CaL sensitivity to b-adrenergic stimulation, which may favour the occurrence of the triggered activity. 153 Hence, the effects of hyperthyroidism on depolarizing currents are incompletely defined.

Repolarizing currents
THs were shown to increase I K1 by activating channel open probability, while the resting membrane potential remained unchanged. 146,148 It was ily T4, which is converted to T 3 , the more metabolically active thyroid hormone, by the enzyme 5 0 -ionidase. Thyroid hormone effects can be genomic or non-genomic. Genomic effects are mediated by binding of T 3 to the nuclear thyroid a-receptor-1 (TRa1), which interacts with the thyroid release elements (TREs) to promote/suppress thyroid hormone-regulated genes. Non-genomic effects are mediated by T 3 and T 4 as they interact with extra-nuclear receptors, which may or may not be structurally related to the thyroid receptor. (B) Hyperthyroidism leads to high-output heart failure (HF), causing left atrial enlargement (LAE), activation of the renin-angiotensin-aldosterone system (RAAS), and increased adrenergic stimulation. Altered intracellular Ca 2þ promotes the formation of early (EADs) and delayed afterdepolarizations (DADs) from the pulmonary veins (PVs). Action potential duration (APD) shortening promotes re-entry. Finally, hypertension (HTN) also contributes to pro-fibrillatory left-atrial structural remodelling. TRH, thyrotropin-releasing hormone; TSH, thyroid-stimulating hormone; DNA, deoxyribonucleic acid.
New aspects of endocrine control of atrial fibrillation and possibilities for clinical translation proposed that the rapid time course of action of T 3 on I K1 is suggestive of a non-genomic mechanisms. 157 THs have been reported to differentially increase ventricular I to without affecting its atrial counterpart. 148,150,152,158 Conversely, Kv1.4 was found to be reduced [159][160][161] and Kv4.2 was unaffected 156,159-161 by THs, while Kv1.5, the I Kur a-subunit, was increased in hyperthyroid animals. 151,156,[159][160][161] APD prolongation due to decreased I Ks was reported in thyroidectomized guinea pigs. 149 Finally, Kv1.2 has been reported to be reduced 159,160 and Kv2.1 unchanged 159,161 or increased 151,160 in hyperthyroid models. Hence, the mechanisms of hyperthyroidism-induced APD shortening are likely multifactorial.
THs also modulate intracellular Ca 2þ homeostasis characterized by decreased phospholamban and increased SERCA2, potentially promoting the occurrence of proarrhythmic afterdepolarizations in hyperthyroid rats. 162 Similarly, pulmonary vein cardiomyocytes from a hyperthyroid rabbits have higher automaticity, more frequent early and delayed afterdepolarizations and shorter APD. 163 The pro-AF changes encountered in hyperthyroidism are summarized in Figure 6.

Thyroid dysfunction and cardiac remodelling
Dysregulated THs mediate multiple cardiac remodelling processes, e.g., necrosis, apoptosis, inflammation, and regression to the foetal heart phenotype. However, cardiac fibrosis remains the hallmark of AF-associated remodelling. THs were shown to downregulate interstitial collagen content 164 and collagen I gene expression in the rat myocardium and cardiofibroblast culture 165 and increased degradation of LV collagen-I/III protein, associated with an activation of MMPs and inhibition of TIMPs, in hyperthyroid rats. 166,167 T3 supplementation in a rat model of ischemia/ reperfusion inhibited TGF-b1, reduced scar size, and improved cardiac performance 168 ; T3 was also shown to inhibit activator-protein-1 (AP-1), 169 involved in the stimulation of MMPs and collagen mRNA. 170 RAS appears to play a role in T4-induced cardiac hypertrophy, 171 which was prevented by treatment with angiotensin-converting enzyme-inhibitor or angiotensin-1 receptor blocker. 172 While hyperthyroidism in rats decreased LV levels of TGF-b1, SMAD2/3, total and phospho-serin368-Cx43 without enhanced interstitial collagen deposition, the TGF-b1 and SMAD2/3 were increased in hypothyroid rats. 173 Although, the majority of studies argue for the anti-fibrotic effects of THs, longstanding hyperthyroidism has been demonstrated to impair LV function and increase interstitial fibrosis in hamsters. 174 Thus, the TH-induced cardiac remodelling may continue to develop over time.
Hypothyroidism is primarily thought to exert profibrotic phenotype associated with increases in the LV TGF-b1 and procollagen-I mRNAs and protein, 164 induced LV hypertrophy with fibrotic lesions, and upregulated a-SMA expression; all these changes were reversed by euthyroid state. Collagen-I/III gene expression was unaltered, while TGF-b1, CTGF, IL-1, and MCP1 gene expression were increased in hypothyroid rats. 175 Interestingly, the TGF-b1 gene promoter has binding sites for Sp1, a critical transcription factor that interacts with TRH-binding protein associated factors, [176][177][178] while pro-a1(I) collagen gene contains a binding site for the receptor, which functions as a TRE. 179 Hypothyroidism was also shown to increase LV collagen-based diastolic wall stiffness 180 and content of collagen and glycosaminoglycans in rat LV. 181 Experiments in cultured rat cardiofibroblasts found increased biosynthesis of fibrillar collagen under TH-depleted conditions 179 and increased proliferation in both TH-depleted 179 and TH-treated cells. 165 Atrial remodelling mediated by thyroid dysfunction is not fully understood; however, a recent study in rats showed that both hypo-and hyperthyroidism increased AF vulnerability, that, as well as the decreased LV and LA dimensions, AERP prolongation, and atrial fibrosis, were decreased by T4 administration. 182 The cross-sectional area and diameter of LA myocytes were reduced in hypothyroid and increased in hyperthyroid rats. 182 While some studies reported an association of hypothyroidism with LA remodelling (marked by increased LA diameter) and increased preoperative AF incidence in patients with heart valvular disease 183 or dilated cardiomyopathy, 184 others did not confirm this. 185,186 The impaired ejection fraction, presence of multiple valvular lesions, and a lower recovery rate of LA enlargement after valve surgery were also observed in hypothyroidism. 183 Overt and subclinical hypothyroidism also increased the risk of post-operative-AF in the patients after cardiac surgery [187][188][189] At the molecular level, hypothyroidism was associated with increased serum levels of CRP, TNF-1a, IL-6, and TGF-b1 in rats, induced secretion of the cardiac stress markers ANP, brain natriuretic peptide (BNP) (regulated by TH) and cardiac troponin-T. 175 Hyperthyroidism also caused increase in cardiac TGF-b1 in cardiac hypertrophy mediated by angiotensin-II receptors 190 and was associated with increased protein and ribosome synthesis. 191,192 These observations suggest that TH are important regulators of cardiac remodelling ( Figure 6).

Calcitonin and AF
Calcitonin (CT) is canonically secreted by parafollicular cells (C cells) of the thyroid gland and is a 32 amino acid single-chain peptide that is cleaved from a precursor pro-CT by protein convertases. 193 Human CT originates from the CT-related polypeptide-alpha (CALCA) gene on chromosome-11 (ID: ENSG00000110680) that also encodes alphacalcitonin gene-related peptide (aCGRP, a potent vasodilator with functions in the nervous and vascular systems). 194 CT plays a well-known role in bone metabolism 195 and plasma Ca 2þhomeostasis. 196 Extra-thyroidal CT expression is present in organs, such as the brain, uterus, prostate, and central nervous system. 197,198 We recently discovered that atrial cardiomyocytes actively produce CT. 199 Regardless of where CT is produced, it exerts its effects via binding to the CT receptor (CTR), 200 the seven-transmembrane domain class II (family B) G protein-coupled receptor that can couple to Gs, Gi, or Gq proteins. 201,202 The CTRs are expressed in tissues such as kidneys, 203 , osteoclasts, 204 skeletal muscle, 205 and recently identified in human atrial fibroblasts. 199 A wide distribution of the CT and CTR indicates that CT-CTR signalling may be involved in the (patho)physiology of multiple systems, including the heart.
The role of the CT-CTR cascade in AF is unclear, however, key risk factors for AF, age 206 and BMI, 207 are associated with decreased circulating CT-levels and CTR single-nucleotide polymorphisms respectively. 208,209 Early work in dog and rabbit models of Ca 2þ -induced arrhythmias observed antiarrhythmic effects of CT on Ca 2þ -induced ventricular arrhythmias 210 and the inhibition of atrial chrono-/inotropic function. 211 The mechanisms of these effects are unknown, though studies in non-cardiac cells showed that CT affected ion fluxes (e.g. neuronal Ca 2þ -currents, 212 kidney Ca 2þ -channels and NCX 213 and, intracellular Ca 2þ214,215 and implicated in AF pathogenesis 216 mitochondrial Ca 2þ influx, 217 in CTsecreting cells) and channel expression (e.g. neuronal Na V 1.3, Na V 1.8, and Na V 1.9 218 ). RA cardiomyocytes from patients with persistent AF secrete six-fold less CT compared to sinus-rhythm controls. 199 of atrial cardiomyocyte-CT in an atrial-specific LKB1-KD model of spontaneous AF caused three-fold higher incidence and 16-fold longer duration of spontaneous AF episodes, which commenced at a younger age vs. control LKB1-KD mice. 199 Overexpression of CT in atrial cardiomyocytes of the LKB1-KD mice prevented atrial arrhythmia. 199 Global deletion of the CTR in mice resulted in increased AF-inducibility. 199 These findings point to a potentially important relationship between the cardiac CT-CTR axis and AF arrhythmogenesis.
CT-CTR signalling is important for tissue fibrogenesis, as it regulates collagen homeostasis, e.g., CT inhibits collagen breakdown in bones 197 and in chondrocytes, 219 while CTR-cKO promotes fibrosis in murine skeletal muscle. 220 Binding of cardiac CT to the surface CTRs of neighbouring human atrial fibroblasts inhibits the production of ECM proteins like collagens-I/III, TIMP2, and SERPIN. 199 In patients with persistent AF, accompanied by significant fibrotic remodelling, fibroblast-CTR localization was altered and confined to the intracellular space, thus, precluding CT from binding CTR ( Figure 7A). CTR-KO mice or atrial-specific cardiomyocyte-CT-KD in the LKB1-KD mice exacerbated atrial fibrosis, suppressed by overexpression of CT in atrial cardiomycytes. 199 In the light of this work ( Figure 7B), maintaining CT-CTR signalling in atrial myocardium may benefit patients with AF.

Towards 'hormonal therapeutics' in AF
The evidence discussed in this review unequivocally demonstrates a potent regulatory role of both endocrine and cardiac para-/autocrine In healthy hearts, atrial cardiomyocytes secrete calcitonin, which binds to the calcitoninreceptors (CTRs) on atrial fibroblasts, controlling extracellular matrix deposition and helping to maintain normal sinus rhythm. In AF, calcitonin signalling is disordered by reduced secretion of calcitonin by atrial cardiomyocytes and reduced calcitonin-receptor responsiveness; these changes impede the calcitonin-mediated brake on fibrogenesis causing atrial fibrosis and increased arrhythmogenesis (created with Biorender.com). (B) CTRs (green) in atrial fibroblasts co-stained with filamin A (red) and 4 0 ,6-diamidino-2-phenylindole (DAPI) (blue). Relocalization of CTRs from the cell membrane to the nucleus explains CTR hyporesponsiveness. Scale bar = 50 lm; adapted from Moreira et al. 199 New aspects of endocrine control of atrial fibrillation and possibilities for clinical translation systems on myocardial function and structure in AF. Although currently not performed, screening of patients with AF for both systemic and cardiac (when possible) hormonal imbalance (in combination with other conventionally used diagnostic tests) may potentially advance treatmentstratification and existing treatment options with the new hormonebased therapies.
Such therapies would first of all aim to treat the prime underlying endocrine pathology (e.g. diabetes, pheochromocytoma, and obesity) with conventional therapies to reduce the risk of new-onset AF or prevent arrhythmia progression. In some cases, available therapies can fully cure the underlying endocrine disorder (e.g. resection of pheochromocytoma) and, hence, reduce risk of AF. Current treatment options for common endocrine conditions (e.g. diabetes and obesity) are often suboptimal; nevertheless, they help to reduce, while not completely eliminating, the risk of AF onset and progression. Furthermore, contemporary medications with improved safety profile may offer improved possibilities for control of AF, for example, with sodium-glucose cotransporter-2 inhibitors that may reduce risk of AF in T2DM, 221 or with glucagon-like peptide-1 receptor agonists (except albiglutide 222 ) that advantageously do not increase risk of AF in obese patients with DM. 223 Looking into the future, control of metabolism with, for example, modified synthetic secreted factors or inhibitors may not only improve cardiometabolic conditions like DM and hypertension 224 but may also hold promise to control AF risks associated with these serious pathologies.
Some endocrine therapies targeting selective pathways, e.g., inflammation and fibrosis, are likely to aid treatment of specific type(s) of AF, e.g., corticosteroids, due to their potent anti-inflammatory properties, might reduce AF recurrence after ablation procedures 225 and incidence of post-operative AF. 226 Selective inhibition of inflammation, for instance with GILZ small molecules, may help to circumvent undesirable sideeffects of broad-spectrum glucocorticoids and improve therapeutic options in AF. 13 Availability of the RAAS antagonists [angiotensin-II-converting enzymes inhibitors and angiotensin II receptor blockers (ARB)], owing to their pronounced antifibrotic effects, may not only control blood pressure in hypertensive patients but to also reduce the occurrence, development, and duration of AF. 227 A recent meta-analysis of 7914 patients showed that aldosterone pathway blockade with MR antagonists limited AF recurrence and, to a lesser extent, prevented the new onset of AF. 228 Since NPs counter-balance RAAS, recombinant human NPs (e.g. nesiritide-recombinant BNP) combined with neprilysin inhibitors (e.g. sacubitril, enhancing NP signalling) and ARBs (e.g. valsartan), denoted ARNi (currently approved for the heart failure management) 228 or synthetic modified NPs designed to preferentially enhance signalling via specific NPRs, may also have potential benefits for AF management.
Sex hormone (testosterone and oestrogen) replacement therapies in patients with AF may also help to prevent AF. 229,230 However, altering sex hormone-dependent pathways may increase the risk of stroke, cardiac arrest, and life-threatening ventricular arrhythmias (due to altered ventricular repolarization and prolonged QT interval). 231,232 Thus, potential risks of such therapies should be carefully balanced against their benefits.
TH replacement therapy can also be beneficial. 233,234 Disrupted CT-CTR signalling in AF might be amenable to CT-based therapies used to treat conditions like osteoporosis and Paget's disease. However, their use is limited due to a release of anti-CT antibodies in some patients and CTR internalization during prolonged CT treatment. 235 Gene-therapy to overexpress CT in atrial cardiomyocytes might offer a tool to manipulate CT levels in a controlled manner. Patients with persistent AF do not maintain membrane CTR localization ( Figure 7A) 199 ; thus, strategies to normalize localization of the CTR is a necessary and challenging objective in attempts to exploit CT-CTR signalling to prevent atrial structural remodelling in AF-patients. In addition, off-target effects of CTRactivation need to be avoided.

Conclusions
It is clear that endocrine/paracrine/autocrine effects play an important role in AF pathogenesis and might present interesting, presently underdeveloped, therapeutic targets. AF management is still very challenging, with many obstacles to optimal management. 236 Recent discoveries, like those of cardiac CT-production and involvement in AF, novel molecular mediators (like GILZ) and the potential mechanistic role of inflammatory signalling, highlights how little we know about endocrine control of AF and how much more there is to learn in order to harness the full therapeutic potential of this critical system.

Data availability
No new data were generated or analysed in support of this research.