The autonomic nervous system and cardiac GLP-1 receptors control heart rate in mice
ABSTRACT
Objectives: Glucagon-like peptide-1 (GLP-1) is secreted from enteroendocrine cells and exerts a broad number of metabolic actions through activation of a single GLP-1 receptor (GLP-1R). The cardiovascular actions of GLP-1 have garnered increasing attention as GLP-1R agonists are used to treat human subjects with diabetes and obesity that may be at increased risk for development of heart disease. Here we studied mechanisms linking GLP-1R activation to control of heart rate (HR) in mice. Methods: The actions of GLP-1R agonists were examined on the control of HR in wild type mice (WT) and in mice with cardiomyocyte-selective disruption of the GLP-1R (Glp1rCM—/—). Complimentary studies examined the effects of GLP-1R agonists in mice co-administered propranolol or atropine. The direct effects of GLP-1R agonism on HR and ventricular developed pressure were examined in isolated perfused mouse hearts ex vivo, and atrial depolarization was quantified in mouse hearts following direct application of liraglutide to perfused atrial preparations ex vivo. Results: Doses of liraglutide and lixisenatide that were equipotent for acute glucose control rapidly increased HR in WT and Glp1rCM—/— mice in vivo. The actions of liraglutide to increase HR were more sustained relative to lixisenatide, and diminished in Glp1rCM—/— mice. The acute chronotropic actions of GLP-1R agonists were attenuated by propranolol but not atropine. Neither native GLP-1 nor lixisenatide increased HR or developed pressure in perfused hearts ex vivo. Moreover, liraglutide had no direct effect on sinoatrial node firing rate in mouse atrial preparations ex vivo. Despite co-localization of HCN4 and GLP-1R in primate hearts, HCN4-directed Cre expression did not attenuate levels of Glp1r mRNA
transcripts, but did reduce atrial Gcgr expression in the mouse heart. Conclusions: GLP-1R agonists increase HR through multiple mechanisms, including regulation of autonomic nervous system function, and activation of the atrial GLP-1R. Surprisingly, the isolated atrial GLP-1R does not transduce a direct chronotropic effect following exposure to GLP- 1R agonists in the intact heart, or isolated atrium, ex vivo. Hence, cardiac GLP-1R circuits controlling HR require neural inputs and do not function in a heart-autonomous manner.
1.INTRODUCTION
Glucagon-like peptide-1 (GLP-1) is a gut hormone synthesized pre- dominantly in enteroendocrine L cells of the small and large intestine. Although GLP-1 levels are low in fasting or interprandial states, GLP-1 secretion and circulating levels of GLP-1 rise rapidly following meal ingestion. Original concepts of GLP-1 action described its role as an incretin hormone that augmented glucose-dependent insulin secretion, via a direct effect on islet b-cells. Subsequent studies extended the islet actions of GLP-1 to encompass stimulation of somatostatin and inhibition of glucagon secretion, mechanisms converging on the reduction of meal-related glycemic excursions. These pancreaticactions of GLP-1 are mediated by a distinct GLP-1 receptor (GLP-1R), expressed on islet b- and d-cells.GLP-1 also exerts pleiotropic actions beyond the islet, consistent with the broad extra-pancreatic expression of the GLP-1 receptor. Notably, GLP-1 inhibits gastric emptying through neural pathways and reduces food intake via activation of a hypothalamic and brainstem network of central nervous system GLP-1 receptors [1]. GLP-1 also promotes natriuresis, reduces blood pressure, and inhibits chylomicron secre- tion, mechanisms mediated through the canonical GLP-1 receptor [2]. The actions of GLP-1 to control blood glucose and body weight are conserved in humans thereby supporting development of multiple GLP- 1 receptor agonists for the treatment of type 2 diabetes (T2D).Although control of blood glucose remains a primary goal in the treatment of T2D, substantial effort is simultaneously directed toward reducing the development of diabetes-associated microvascular and macrovascular complications. Notably, reduction of blood pressure, control of dyslipidemia, and judicious use of anti-platelet agents represent important complimentary goals for management of T2D [3].
The introduction of mandatory cardiovascular safety studies by regu- latory authorities has further heightened the interest in whether glucose-lowering agents exert independent actions that modify car-diovascular risk in subjects with nascent or established cardiovascular disease. In this regard, the SGLT-2 inhibitor empagliflozin was recently shown to reduce rates of cardiovascular death in a large cardiovascular outcome study [4], further elevating the importance of understanding the non-glycemic actions of drugs used to treat T2D.One of the first non-glycemic actions described for native GLP-1 and degradation-resistant GLP-1R agonists was a rapid increase in heartrate (HR) [5,6]. The increase in HR is mediated through the canonical GLP-1 receptor [6,7], independent of changes in blood glucose, and associated with activation of autonomic sympathetic preganglionic brainstem neurons in normoglycemic rats [7,8]. GLP-1 may also modify HR through attenuation of parasympathetic tone [9], however the relative contribution of sympathetic vs. parasympathetic tone to GLP-1R-dependent control of HR remains uncertain [10,11].The identification of GLP-1R expression in the atria of mice [12,13] and in the sinoatrial node of monkey [14] raised the additional possibility thatGLP-1 may also modulate HR via a direct effect through the cardiac GLP- 1R. Further support for this hypothesis derives from observations that mice with cardiac-selective reduction of GLP-1R expression (Glp1rCM—/—) exhibit a reduction in basal HR assessed over 24 h [15].
Hence, the available data suggest that GLP-1 may control HR indirectly, through modulation of autonomic nervous system activity, as well as directly, through control of pacemaker activity via the atrial GLP-1R.To elucidate the relative importance of and inter-dependence of these pathways, we have now studied the regulation of HR in control and Glp1rCM—/— mice treated with lixisenatide or liraglutide, GLP-1R pep-tide agonists with distinct pharmacokinetic profiles. Notably, lix-isenatide is a short-acting exendin-4-derived GLP-1R agonist, whereasthe human GLP-1 analog liraglutide, exhibits a more protracted duration of action, via acylation and non-covalent association with albumin [16]. The importance of the autonomic nervous system was examined in mice co-administered propranolol or atropine. Compli- mentary experiments examined HR responses to GLP-1R agonists inisolated perfused mouse hearts under aerobic and ischemia/reperfu- sion conditions, and in atrial preparations ex vivo. Our findings reveal temporally distinct contributions from both the sympathetic nervous system and cardiac GLP-1Rs in the HR response to GLP-1R agonism in mice in vivo. In contrast, the isolated perfused mouse heart ordissected atrial preparations did not exhibit a direct chronotropic response to GLP-1R agonists ex vivo. Hence, the atrial GLP-1R me- diates a HR response to GLP-1R agonism in the context of the intact mouse heart in vivo, but is not sufficient for transduction of a heart- autonomous chronotropic response ex vivo.
2.MATERIALS AND METHODS
All mice used in these studies were adult males housed under pathogen-free conditions in microisolator cages and maintained on a 12-h light/dark cycle with free access to standard rodent diet (#2018, 18% kcal from fat; Harlan Teklad) and water, unless otherwise noted. All experiments were carried out in accordance with protocolsapproved by the Animal Care Committee at Mt. Sinai Hospital and the Toronto Centre for Phenogenomics and were consistent with ARRIVE guidelines. Glp1rCM—/— mice and Cre WT controls were generated as described [15]. Hcn4-Cre mice were provided by Dr. Andreas Ludwig. The generation of these mice and tamoxifen induction of Cre expres- sion were described previously [17]. Glp1rSAN—/— and GcgrSAN—/—mice and controls (wild type, Cre, floxed) were produced by breedingHcn4-Cre hemizygous mice with floxed Glp1r mice [18]or floxed Gcgrmice [19], respectively. Wild type C57BL/6J mice (used for experi-ments in Figures 3, 4 and 6) were purchased from The Jackson Laboratory (Bar Harbor, ME).Mice were fasted for 6 h (7ame1pm) and then given a single ip in- jection of vehicle (PBS) or lixisenatide immediately prior to receiving oral glucose (1.5 g/kg body weight) by gavage. Liraglutide was administered as a single ip injection 2 h prior to oral glucose challenge. Blood glucose levels were measured in tail vein blood samples at time 0 and for up to 120 min following oral glucose administration using a hand-held glucose meter (Contour, Bayer Inc., Mississauga, ON).Heart rate was measured in conscious, freely moving mice using an implanted radiotelemetry device (PA-C10, Data Sciences International, St. Paul, MN). All mice were allowed a minimum period of 1 week to recoverfrom device implantation surgery prior to initiation of data collection. Mice were given ip injections of vehicle, lixisenatide (10 mg/kg; Sanofi), lir- aglutide (30 mg/kg; Novo Nordisk), propranolol hydrochloride (5 mg/kg;Sigma Aldrich, Oakville, ON), or atropine methyl nitrate (2 mg/kg; Sigma Aldrich, Oakville, ON). Mice that received liraglutide were not given another injection until 48 h later, to ensure drug wash out.
All injections were administered at approximately 4pm each day.Nucleic acid isolation and analysis of mRNA expression Genomic DNA was isolated from mouse tissue using a DNeasy Bloodand Tissue Kit (Qiagen, Toronto, ON). PCR amplification of genomic DNA sequences flanking exons 6 and 7 of the Glp1r were generated using the primer pair 50-GGA TCC GAA CTG AGG TCC TC-30 and 50-GGG GTG ATA TTT GGC CAT ATG AG-3.’Total RNA was extracted from tissues using Tri Reagent (Molecular Research Center Inc., Cincinnati, OH). cDNA was synthesized from DNase I-treated (Thermo-Fisher Scientific, Markham, ON) total RNAusing random hexamers and Superscript III (Thermo-Fisher Scientific,Markham, ON). Real-time PCR was carried out using a QuantStudio 5System and TaqMan Gene Expression Assays (Thermo-Fisher Scien- tific, Markham, ON) for Glp1r (Exon 5-6; Mm00445292_m1) or Gcgr (Exon 7-8; Mm00433550_g1). Relative mRNA levels were quantifiedusing the 2—DCt method and cyclophilin (Ppia) mRNA expression fornormalization. Cre cDNA was amplified by PCR using the primer pair 50-CCC GCG CTG GAG TTT CAA TA-30 and 50-CTT CGC CCA GTT GATCAT GTG-3.0 Glp1r Exons 5-8 cDNA was amplified by PCR using the primer pair 50-CCT GAG GAA CAG CTC CTG TC-30 and 50-ACA ATC CCCCAT GGG ATA AC-3.0 Gcgr Exons 5-13 cDNA was amplified by PCR using the primer pair 50-AGC ACC GCC TAG TGT TCA AG-30 and 50-AAA CAG TAG AGA ACA GCC ACC A-3.0Hearts were removed from mice, cannulated through the aorta, and perfused at constant pressure using a Langendorff apparatus as described [20]. Continuous pressure traces were recorded from aballoon placed in the left ventricle, and all data analysis is based on the 5 min average for a given parameter. For aerobic perfusion experi- ments, heart rate and left ventricular developed pressure (LVDP) were assessed in response to increasing doses of lixisenatide (0.5e50 nM;Sanofi) or vehicle. Isoproterenol (100 nM; Sigma Aldrich, Oakville, ON) was used as a positive control at the end of each perfusion. For ischemia/reperfusion injury experiments, heart rate and LVDP were determined at baseline (30e40 min aerobic perfusion; analysis is shown for the final 10 min of baseline data prior to ischemia), during 30 min of no-flow global ischemia, and during 40 min of reperfusion.
GLP-1 (0.5 nM) orlixisenatide (5 nM) was delivered via a syringe pump into a warmed bubble trap situated directly above the perfused heart. Drug dose cal- culations were based on an approximation of 3 ml/min total flow through the heart. Ischemia preconditioning was performed before the 30 min ischemic period by 4 cycles of 5 min ischemia, 5 min reperfusion.Atrial preparations were from 12- to 16-wk-old C57BL/6 mice. Hep- arinized mice were sacrificed and live atria dissected in Tyrode’s so- lution (35 ◦C) consisting of (in mmol/L) 140 NaCl, 5.4 KCl, 1.2 KH2PO4,1.0 MgCl2, 1.8 CaCl2, 5.55 glucose, and 5 HEPES, with pH adjusted to7.4 with NaOH. Atria were pinned to the bottom of a Sylgard-coated 30-mm petri dish in “butterfly” fashion so that the crista terminalis could be identified and used as a landmark for sinoatrial node position. Atria were incubated in 1 mM di-4-ANEPPS (Molecular Probes, Eugene,OR), a voltage sensitive dye, for 10 min followed by superfusion with 37 ◦C Krebs solution of the following composition: 118.0 mM NaCl,4.7 mM KCl, 1.2 mM KH2PO4, 1.2 mM MgSO4, 1.0 mM CaCl2,25.0 mM NaHCO3, 11.1 mM glucose, at a constant flow rate of w10 ml/min. The superfusate was continuously bubbled with 95% O2e5% CO2 in a 37 ◦C water bath, resulting in pH 7.4. Drugs to be tested for their effect on sinoatrial node (SAN) firing rate were added directly to the warmed oxygenated superfusate and allowed to flowdirectly onto the pinned atrial tissue. Each treatment was applied for 3 min and heart rate values were determined from an area of interest within the sinoatrial node. The integrity of the SAN, and its respon- siveness to chronotropic agents was routinely tested at the end of anexperiment by the addition of 100 nM isoproterenol. 100% of the atria tested showed a w200 b.p.m. increase in their SAN firing rates in response to isoproterenol. The imaging system consisted of illumi- nating light provided by a 120-W quartz mercury lamp light source (X- Cite exacte, EXFO Life Sciences, Mississauga, ON) through a band- pass filter and off a dichroic mirror (Olympus).
A shutter (EXFO Life Sciences, Mississauga, ON) was used to minimize tissue exposure to excess light (only open during image acquisition). The fluorescent lightemitted from the preparation passes through the dichroic mirror and is further long-pass (>590 nm) filtered using a Schott glass filter (Melles Griot, Ottawa, ON) before reaching the camera. A charge-coupled device camera (model Cascade 128 , Photometrics, Tucson, AZ) was used in binning mode (64 64 pixels, 16-bit resolution, 1000 frames/s). In the present application, it was equipped with a 25-mm focal length lens (Computar, Commack, NY) and a 5-mm spacer, resulting in a 10 10 mm field of view (167 167 mm/pixel). Image acquisition software ImagePro Plus 5.1.2.59 (Media Cybernetics,Bethesda, MD) and the camera were connected via an image acqui- sition board (model PCI- 1422, National Instruments, Austin, TX) to a personal computer. All data were processed offline using ImagePro Plus 5.1.2.59 software.As a control for the lack of actions of liraglutide in isolated atria, the same stocks of liraglutide were used to assess heart rates in 12e16wk old anesthetized C57BL/6 mice via implanted Millar 1.4 French Millar blood pressure probe passed through the right common carotid artery into the ascending aorta. Liraglutide or vehicle (5 ml of 0.9% saline) was administered intravenously through a polyethylene catheter (PE10) in the left jugular vein, at a rate of1.5 ml/s and heart rate was recorded for 5 min before addition of next dose. All experiments were performed with experimenters blinded to the drug treatment group until after data analysis was complete.Results are presented as mean SE. Statistical significance was determined by two-tailed Student’s t-test or one- or two-way ANOVA with Bonferroni post hoc analysis using GraphPad Prism version 5.02 software (San Diego, CA). A p value < 0.05 was considered statisti- cally significant. 3.RESULTS To assess and compare the actions of a short-acting GLP-1R agonist, lixisenatide, and a longer-acting GLP-1R agonist, liraglutide, we first carried out dose-ranging studies to identify doses of these peptides that produced comparable glucose-lowering activity in mice in vivo. Notably, acute administration of 10 mg/kg lixisenatide and 30 mg/kg liraglutide produced comparable glycemic reductions in response to oral glucose (Figure 1A and B) in both control (a-myosin heavy chain promoter-Cre (aMHC-Cre; Cre WT) mice [15] and in mice with cardiomyocyte-selective inactivation of the GLP-1 receptor Glp1rCM—/—). Hence, we selected these two pharmacodynamically equivalentglucoregulatory doses for further analysis of chronotropic activity.We next assessed the actions of liraglutide, an extensively studiedGLP-1R agonist used in the treatment of diabetes and obesity, on HR in mice. HR was increased for a prolonged period of time, remaining significantly elevated at 16e18 h following a single injection of lir- aglutide (Figure 2A and Supplementary Figure 1). Furthermore, the actions of liraglutide to increase HR were attenuated in Glp1rCM—/—mice (Figure 2B and Supplementary Figure 1), extending previousfindings implicating a role for the cardiomyocyte GLP-1 receptor in the control of basal HR [15]. Consistent with these findings, HR trended higher following lixisenatide administration in WT mice and to a lesser extent in Glp1rCM—/— mice (Supplementary Figure 2).As GLP-1R agonists increase activation of autonomic catecholamin- ergic neurons in the rat central nervous system [7,8], we next ascertained whether the acute chronotropic effects of lixisenatide or liraglutide were modified by pharmacological blockade using the b- adrenergic antagonist propranolol. Administration of propranololcompletely attenuated the acute increase in HR (from 1 to 6 h) observed with lixisenatide (Figure 3A). Although GLP-1R agonists are thought to increase HR in part through inhibition of parasympathetic nervous system activity, the muscarinic receptor antagonist atropinedid not eliminate or potentiate the acute lixisenatide-mediated induc- tion of HR over the first 3 h (Figure 3B). In contrast, HR was no longer increased (in the presence of co-administered atropine) from 3 to 6 h after lixisenatide administration (Figure 3B). Propranolol similarlyattenuated the acute (0e3 h) induction of HR following liraglutide administration, however a very modest but detectable increase in HR remained evident in mice treated with both liraglutide and propranolol when analyzed from 3 to 6 h after liraglutide injection (Figure 4A). Furthermore, liraglutide continued to increase HR in atropine-treated mice, when analyzed from 1 to 3 h or 3e6 h after liraglutideadministration (Figure 4B). Taken together, these findings demonstrate independent, temporally distinct contributions from the sympatheticnervous system and the cardiac GLP-1R in the chronotropic response to GLP-1R agonism in mice.Immunohistochemical analysis of GLP-1 receptor expression in the monkey heart co-localized the sinoatrial node (SAN) GLP-1R with hyperpolarization-activated cyclic nucleotide-gated channel 4 (HCN4), a major ion channel regulating pacemaker function [14]. Accordingly, we hypothesized that mice expressing Cre recombinase under the control of the HCN4 promoter [17] would enable elimination of sino- atrial Glp1r expression in the mouse heart. To this end, we matedHcn4Cre mice with floxed Glp1rFl/Fl mice to generate Glp1rSAN—/— mice. Unexpectedly, we did not detect reduction of right atrial Glp1r expression (or transcript length) in Glp1rSAN—/— mice (Figure 5A and Supplementary Figures 3 and 4), despite detectable Cre mRNAexpression and recombination of Glp1r genomic DNA in the right atrium (Supplementary Figures 3e5). As the structurally related glucagon receptor (Gcgr) has been localized to the mouse SAN [21] and like GLP-1, transduces an acute increase in HR [22], we examined whether Hcn4Cre mice would enable reduction of atrial Gcgr expres-sion. In contrast to findings in Glp1rSAN—/— mice, GcgrSAN—/— mice exhibited a reduction in both the expression and length of Gcgr mRNAtranscripts in the right atrium (Figure 5B and Supplementary Figures 6 and 7). Hence, the failure to achieve reduction of atrial Glp1r mRNA transcripts in Glp1rSAN—/— mice (Figure 5A) is consistent with lack of co-expression of HCN4 and GLP-1R in the mouse heart.Although endogenous GLP-1 exerts glucoregulatory actions via both neural circuits and activation of the b-cell GLP-1R, GLP-1R agonistsrobustly stimulate insulin secretion in isolated islets ex vivo. To determine whether the atrial GLP-1R was sufficient for GLP-1R- dependent activation of HR in a heart-autonomous manner, we assessed HR in isolated perfused mouse hearts ex vivo. Neither native GLP-1 nor lixisenatide increased HR under normoxic conditions (Figure 6A); however, HR was robustly increased by the b-adrenergicagonist isoproterenol in the same heart preparations ex vivo (Figure 6A). Similarly, we were unable to detect an effect of GLP-1 or lixisenatide on recovery of left ventricular developed pressure (LVDP) following induction of global no flow ischemia in isolated hearts ex vivo (Figure 6B), whereas ischemic preconditioning robustlyincreased recovery of LVDP in similar experiments. Moreover, increasing doses of lixisenatide (0.5e50 nM) had no effect on LVDP in aerobic perfused mouse hearts ex vivo (Figure 6A). Hence, although the atrial GLP-1R is required for GLP-1R-dependent control of heartrate in vivo, it is not sufficient for transduction of a GLP-1R-dependent signal to increase HR or ventricular function in denervated perfusedhearts ex vivo.To determine whether GLP-1R agonists such as liraglutide directly regulate atrial depolarization, we assessed frequency of the SAN firing rate in isolated mouse atrial preparations ex vivo. Direct application of saline or liraglutide to perfused atria had no measurable effect on SAN firing rates (Figure 7A). In contrast, administration of liraglutide (fromthe same stock solution) rapidly increased HR (relative to salinevehicle) in intact anesthetized mice (Figure 7B). Hence, GLP-1R ago- nists increase HR through both the autonomic nervous system and the atrial GLP-1R in vivo, but are unable to increase HR through the atrial GLP-1R in a heart-autonomous manner ex vivo. 4.DISCUSSION The development of glucose-lowering agents for the treatment of T2D is now associated with increased scrutiny as to whether and how these drugs independently modulate the risk of developing cardiovascular disease [23]. The non-glycemic actions of GLP-1R agonists encompass reduction of blood pressure and postprandial triglycerides, weight loss, and reduction of inflammation, actions predicted to mitigate the risk of developing cardiovascular compli- cations [2]. Although short-acting drugs such as exenatide and lix- isenatide have been associated with modest and transient increases in HR of several beats per minute (bpm) [24,25], longer-acting agents such as dulaglutide and liraglutide are more potent in vivo and produce greater and more sustained increases in HR up to 10 bpm [26e28]. As increases in HR may enhance the susceptibility to tachyarrhythmia in at risk patients, understanding the mecha- nisms through which structurally distinct GLP 1R agonists control HR may have clinical relevance. Here we show that HR is acutely increased following administration of lixisenatide or liraglutide, in control and Glp1rCM—/— mice. Moreover, the increase in HR is abolished by co-administration of the b adrenergic antagonist propranolol and not further enhanced by attenuation of cholinergic signaling using the muscarinic receptor antagonist atropine. These findings demonstrate that the cardiac GLP- 1R is not essential for the acute chronotropic response to GLP-1R agonists and are consistent with previous studies demonstrating GLP-1R dependent induction of central and peripheral catecholamin- ergic pathways in rats and mice [7,8]. Moreover, our findings are in agreement with data demonstrating attenuation of the acute chrono- tropic response to exendin-4 by propranolol in rats [29]. GLP-1R agonism also potentiated plasma catecholamine responses to exercise in human subjects [30]. Hence the available evidence links acute GLP-1 receptor activation to catecholaminergic activation and induc- tion of HR in vivo.Although the cardiac GLP-1R was not required for the acute induction of HR by lixisenatide or liraglutide, the sustained induction of HR by liraglutide was attenuated in Glp1rCM—/— mice. It seems likely that once the robust yet transient induction of catecholaminergic signaling wanes, the contribution of intrinsic GLP-1R signaling to the induction of HR becomes evident. Indeed a functional role for the GLP-1R in atrial pacemaker cells was indirectly suggested by Pyke and colleagues who co-localized HCN, and GLP-1R expression within the monkey sinoatrial node [14]. Surprisingly, however, we were unable to attenuate atrial Glp1r expression through generation of Glp1rSAN—/— mice, whereas GcgrSAN—/— mice exhibit a significant reduction of atrial Gcgr expression. These findings indicate that the organization of cellular atrial GLP- 1R expression may be species-specific, and the localization of cardiac GLP-1R expression may have evolved to serve a different subset of functions in mice vs. primates. Several lines of evidence support the existence of the GLP-1R within cardiomyocytes in the mouse atria. First, we have detected expression of full length Glp1r mRNA transcripts in atria [31], and RNA levels of the full length atrial Glp1r were markedly reduced in floxed Glp1r mice, following recombination of the Glp1r allele using Cre under the control of the myosin heavy chain promoter [15]. In addition, we observed robust atrial natriuretic factor secretion in mice [12,27] but not in humans [27], following administration of GLP-1R agonists. Moreover, GLP-1R agonists rapidly and directly increased cyclic AMP formation and atrial natriuretic factor secretion in isolated mouse atrial car- diomyocyte cultures ex vivo [12]. In contrast to intense GLP-1R expression in the monkey sinoatrial node [14], Richards and col- leagues detected only a few scattered random atrial cells expressing a reporter gene under the control of the endogenous Glp1r promoter in transgenic mice [13] and single cell RNASeq has demonstrated low level but detectable expression of the Glp1r mRNA in mouse car- diomyocytes [32]. Nevertheless, the precise cellular identity of the repertoire of GLP-1R cells in mouse atria remains incompletely defined and it remains possible that the GLP-1R may also be expressed in non-cardiomyocyte cell types. Taken together, these disparate findings of atrial GLP-1R expression and function highlight the importance of further studies examining whether species-specific patterns of GLP-1R expression and function in the heart might confound interpretation of the cardiovascular actions of GLP-1 in different experimental settings. Our current findings also indirectly establish the importance of neural and/or vascular inputs essential for coupling atrial GLP-1R activation toan increase in HR. Notably, we were unable to detect an increase in HR using native GLP-1 or lixisenatide in isolated perfused hearts. Furthermore, we failed to demonstrate direct GLP-1R-dependentactivation of atrial firing in the mouse atria using preparations of exposed mouse atria ex vivo. In contrast, Glp1rCM—/— mice exhibit reductions in basal HR [15] and attenuated HR responses to liraglutide in vivo. Collectively, these findings imply that although the atrial GLP- 1R is essential for HR regulation in mice in vivo, it is unable to activatea signal linked to pacemaker function and HR ex vivo in the absence of intact neurovascular input.Our study has several limitations. First, we studied the short term effects of GLP-1R activation; however, the relative quantitative importance of the autonomic nervous system vs. the cardiac GLP-1R for the chronic effects of GLP-1R agonists cannot be inferred from our experiments. Second, we used healthy non-obese, normoglycemic mice in our experiments, whereas GLP-1R agonists are used clinically in human subjects with obesity and/or T2D. Third, we did not study the chronotropic actions of GLP-1R agonists in animals with heart disease, or in mouse models with experimental vagotomy or autonomicneuropathy, conditions that may modify the contributions of the cardiac GLP-1R vs the autonomic nervous system to control of HR. Finally, it remains possible that experiments using different concentrations of pharmacological antagonists to achieve parasympathetic or sympa- thetic nervous system blockade over different time periods, may produce different conclusions.Nevertheless, our findings clearly extend current concepts defining how GLP-1R agonists acutely increase HR, integrating contributionsfrom both catecholaminergic pathways and the cardiac GLP-1R (Figure 8). Moreover, our genetic data suggest that, unlike data re- ported for monkeys, atrial GLP-1R expression does not occur in the majority of HCN4 cardiomyocytes. Finally, the importance of the cardiac GLP-1R for HR control is evident only in the living mouse, and is no longer evident when the atrial GLP-1R is activated in the isolatedmouse heart or atrial preparation ex vivo. Collectively, these findings add to our understanding of the cellular sites and pathways linkingGLP-1 receptor activation to the control of HR and cardiovascular Lixisenatide function.