Our study examined male mice because male animals exhibited less variability in myocardial infarction phenotypes and showed more significant resilience to arrhythmia after treatment with GLP-1R/GABA_BR agonists. All animal experiment procedures followed the principles of the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH publication no. 85-23, revised 1996) and were authorized by the Animal Care and Use Committee of Renmin Hospital of Wuhan University under the approval number: 20220301A. All male Sprague–Dawley rats (8–9 weeks old) and male C57BL/6J mice (8–10 weeks old) were purchased from Shulaibao (Wuhan) Biotechnology Co., Ltd. Gabbr1 flox mice (#S-CKO-11747) were on a C57B/J background. The myocardial infarction models were induced by 6-0 suture ligation of the left anterior descending branch of the coronary artery. Sham surgery was performed in the same procedure, except for artery ligation. The success in myocardial infarction modeling is indicated by ST segment elevation from ECG. Semaglutide was injected subcutaneously at a dose of 100 μg/kg/day and treated once a day. Stellate ganglionectomy was performed on all mice under a surgical microscope, following the procedure previously reported (Hill et al., 2024). Lungs were pulled caudally to visualize stellate ganglion between the first and second rib beneath the parietal pleura. Chemical denervation was achieved via 100 mg/kg 6-hydroxydopamine (6-OHDA, diluted into 0.3% ascorbic acid) injection. Baclofen (15 mg/kg) dissolved in 0.9% saline was applied interperitoneally 1 day before semaglutide treatment. Mice were fed for another 1 week after surgery. All mice were fed in a standard environment with controlled light/dark cycles (12-h light/12-h dark), ambient temperature, and humidity. Tail-vein injection of AAV9-cTnT-Cre virus (WZ Biosciences, Inc, Shandong, China, 0.5 × 10^E11 GC/pup) was performed to achieve cardiomyocyte-specific knockout of Gabbr1. All mice were grouped and sacrificed randomly, but no blind test was conducted in the animal experimental procedures.

iPSC-derived cardiomyocyte was obtained from CardioEasy^® (CA2201106, CellAPY, China). The cells were cultured in DMEM/F-12 (11554546, Gibco, United States), 2% B27 insulin-free (A1895601, Thermo Fisher Scientific, United States) penicillin–streptomycin (1%; 100 units/mL penicillin and 100 μg/mL streptomycin) (Cellclone; Genetix Biotech Asia Pvt. Ltd.). The cell culture incubator (Forma™ Steri-Cycle™, 370, Thermo Fisher, USA) was in a humidified air containing 5% CO₂ at 37 °C. A total of 1×10⁶ cells were seeded in a T-25 culture flask (Eppendorf, Hamburg, Germany). Each plate was seeded with an equal amount of cells.

The cardiac function of mice was evaluated at 1 week after surgery with echocardiography (Visual et al., 2100, Toronto, Canada), equipped with a 23-MHz line array transducer. The mice were maintained under 0.5% anesthesia and placed in the supine position on a 37 °C heating pad. M-mode images were obtained to measure left ventricular end-systolic volume (LVESV) and left ventricular end-diastolic volume (LVEDV). Left ventricular ejection fraction (LVEF) and fractional shortening (FS) were obtained from the VEVO system.

The hearts were fixed in 4% paraformaldehyde and sectioned into 5 μm-thick slices. Based on the standard procedure, the hematoxylin–eosin (HE) staining and Masson’s staining in cross-section were performed to evaluate myocardial condition and fibrosis condensation. ImageJ (Fiji) was used for the calculation.

The ventricular sections were cut into approximately 5-µm slices, followed by paraffinization, rehydration, heat-mediated antigen retrieval, and treatment with 3% H₂O₂. After blocking for 1 h with 5% bovine serum, the slices were incubated with the primary antibody overnight at 4 C. Antibody information was as follows: anti-tyrosine hydroxylase (Abcam, ab6211, Germany, 1:500) and anti-cTnT (Abcam, ab8295, Germany, 1:1000). The next day, sections were incubated with the horseradish peroxidase (HRP)-labeled secondary antibody for 2 h at 37 C. Immunofluorescence images were captured using a confocal laser scanning microscope (ZEISS LSM 800).

Atrial pacing was produced through transesophageal programmed electrical stimulation. Mouse anesthesia was maintained by 1% isoflurane. Standard surface ECG was recorded using the PowerLab System (AD instruments) with a subcutaneous ECG surface (lead II). The tracheal tube was inserted into the trachea through the glottis, and the chest fluctuation of the mice was observed to be consistent with the ventilator frequency, which proved that the tracheal intubation was successful. Then, a 2.2F six-polar catheter was inserted into the esophagus near the left atrium, and correct placement was confirmed through burst waves. To correct for the HR, Bazett’s formula-corrected QT interval (QTc) was used. The ECG of mice at the rest state was recorded consecutively for 5–10 min. Subsequently, sodium pentobarbital (50 mg/kg, intraperitoneal) was used to anesthetize animals. With a platinum MAP electrode and stimulation procedures, the monophasic action potentials (MAPs) of the left ventricle were recorded. The paired platinum-stimulating electrode was positioned on the basal surface of the right ventricle to deliver regular pacing. The heart was stimulated with a regular pacing cycle length (PCL). Action potential duration 90 (APD₉₀) was defined as the average repolarization time of 90% of 6–8 consecutive MAPs when the PCL was 150 ms. S1–S1 pacing was used to measure APD and activation latency time (ALT). To induce ALT, PCL was decreased, starting at 150 ms and gradually reduced by 10 ms, and then by 5 ms from 100 to 50 ms, until APD alternans occurred. Ventricular arrhythmias (VAs) were induced by burst pacing with 2 ms pulses delivered at 50 Hz for 2 s, repeated 20 times and separated by 2-s intervals. VA is defined as ventricular tachycardia (VT) or ventricular fibrillation (VF) lasting 2 s or more.

Cardiac samples were homogenized in a tissue lysis buffer (pH: 7.4 with 150 mmol/L NaCl, 1% Triton ×100, and proteinase inhibitor) and centrifuged (10,000 g, 10 min, 4 °C). The supernatant was then collected, and its protein concentration was adjusted to 500 μg/μL. Commercial ELISA kits were used to detect the norepinephrine level (ab287789, Abcam, USA).

A soft glass capillary pulled to a tip resistance of 1.5–2 MΩ (Sutter Instruments, Novato, CA) was used for whole-cell patch clamp. Signals were recorded using an Axopatch 200A (Axon Instruments, Foster City, CA) with a computer-interacted 125-kHz Labmaster board (Axon Instruments). Membrane currents were sampled at 1–2 kHz and filtered at 2 kHz. Series resistance (R _s = 7.0 ± 1.0 MΩ) was compensated by ≈ 80%. Cells to be tested were incubated at 35 °C in the bath solution supplemented with 300 ng/mL pertussis toxin for 4 h. The buffer was supplemented with 2 mM Co²⁺ and 3 mM 4-aminopyridine to block transient rectifying current and calcium current. Voltage commands, data acquisition, and analysis were performed using pClamp 6.0.

The tissue samples were lysed in 1× RIPA buffer (G2002, ServiceBio, Wuhan, China). Subsequently, protein samples were separated using SDS-PAGE and then transferred into the PVDF membrane (IPVH00010, Merk Millipore, Germany). After the PVDF membranes were blocked with 5% fat-free bovine milk for 2 h, specific primary antibodies were incubated overnight at 4 °C. The primary antibodies included. The next day, the membranes were incubated with HRP-conjugated goat anti-rabbit secondary antibodies (ServiceBio, Wuhan, China) for 1 h. Finally, protein bands were visualized through enhanced chemiluminescence (BL523B, Biosharp, China). Antibody information was as follows: anti-αSMA (1:1000, Cell Signaling Technology, United Kingdom, #19245) and anti-TGF-β (1:1000, Cell Signaling Technology, United Kingdom, #3711).

All statistical analyses were performed using GraphPad Prism 8.0 (Inc., La Jolla, CA, USA). Data are presented as mean ± SD. Statistical analyses used repeated two-way ANOVA and paired Student’s *t*-tests where appropriate (*p* <0.05). Values were considered statistically significant when P < 0.05.

GLP-1R is highly enriched in autonomous neurons and cardiomyocytes. To investigate whether the autonomous nervous system is sufficient for mediating GLP-1RA-related fast pacing effect, we blocked cardiac sympathetic innervation via 6-OHDA infusion and surgical sympathectomy. 6-OHDA, a neurotoxic agent that degrades sympathetic nerve terminals, reduced nearly 75% of tyrosine hydroxylase (TH, a key enzyme involved in the production of norepinephrine) fluorescence signals in cardiac tissue, and bilateral stellate sympathectomy depleted 87% of TH⁺ signals (Figures 1A,B). The indicator of sympathetic innervation, norepinephrine, was also significantly reduced after both 6-OHDA infusion and sympathectomy, as determined by ELISA (Figure 1C). Semaglutide treatment increased the HR to a new baseline, whereas heart denervation did not lead to further change in the HR, which suggests that the autonomous nervous system might not be sufficient for GLP-1RA’s positive chronotropic effect (Figures 1D,E).

GABA_BR was identified as a direct binding partner of GLP-1R in a previous biotin proximity labeling screen (Dang et al., 2025). To validate the function of the GABA_BR/GLP-1R interaction, pharmacological activation of GABA_BR with baclofen, a GABA-mimetic GABA_BR agonist, prior to MI induction, reduced the chronotropic effect of semaglutide under physiological conditions (Figures 2A–D). However, the activation of GABA_BR did not accelerate the HR under physiological conditions (Supplementary Figure S1). Moreover, this antagonistic effect of GLP-1RA and GABA_BR on the HR was preserved in post-MI mice. Here, we hypothesize that GABA_BR serves as an antagonizing mediator for GLP-1R regarding HR modulation, potentially through direct receptor cross-talk or intracellular downstream.

To determine whether the baclofen-induced reduction in the HR depends on GLP-1R signaling, we generated cardiac GABABR-knockout (Gabbr1 ^cKO) mice by injecting AAV9-cTnT-Cre into Gabbr1^f/f mice and assessed cardiac electrophysiology in vivo (Supplementary Figure S2). Gabbr1 encodes an essential subunit required for GABA_BR activation. As shown in Figures 3A,B, dual treatment with GLP-1RA and baclofen significantly shortened the action potential duration at APD90 (MAPs, 150 ms) in wild-type mice. This effect was abolished in Gabbr1 ^cKO mice and in those receiving monotherapy with either semaglutide or baclofen (Figures 3G,H). We next evaluated the ALT in the left ventricle by S1–S1 pacing. In Gabbr1-intact mice (Figures 3C,D), combined semaglutide and baclofen treatment attenuated MI-induced ALT prolongation. In contrast, no changes in ALT were observed in Gabbr1 ^cKO mice under either monotherapy or combination therapy (Figures 3I,J). To further assess arrhythmic susceptibility, we performed burst pacing (Figures 3E,F). In Gabbr1^f/f mice, co-administration of GLP-1RA and baclofen reduced both the incidence and duration of post-MI ventricular arrhythmias. However, in Gabbr1 ^cKO mice, dual GLP-1R/GABA_BR activation failed to confer anti-arrhythmic protection (Figures 3K,L). Collectively, these findings demonstrate that GABA_BR activation is required for the anti-arrhythmic effects of GLP-1RAs.

To determine whether dual activation of GABABR and GLP-1R further improves cardiac function, we performed echocardiography in Gabbr1 ^f/f and Gabbr1 ^cKO mice after MI. In wild-type mice, GLP-1RA administration improved post-MI cardiac performance, as reflected by increased LVEF and LVFS and reduced LVESV and LVEDV (Figures 4A,C). In contrast, Gabbr1 ^cKO mice failed to show any improvement in cardiac function or survival following GLP-1RA treatment (Figures 4B,D). Moreover, GLP-1RA reduced the 30-day post-MI mortality rate in wild-type mice, whereas this benefit was abolished in Gabbr1 ^cKO mice (Figures 4E,F). Together, these results indicate that GABABR activation is essential for the cardioprotective and survival benefits of GLP-1RA therapy after MI.

To further assess the role of GABA_BR/GLP-1R dual activation in post-infarction remodeling, we examined cardiac fibrosis using Masson’s trichrome and HE staining. Gabbr1 deficiency did not alter the cardiac morphology or structure after MI, regardless of treatment with semaglutide alone or in combination with baclofen (Figures 5A,B). Quantitative analysis showed no significant differences in collagen deposition or infarct size with GLP-1RA treatment, either alone or with GABA_BR co-activation (Figures 5C,D), indicating that neither GLP-1R nor GABA_BR activation confers acute antifibrotic benefit. Consistently, Western blotting of left ventricular tissue revealed no changes in α-smooth muscle actin (α-SMA, a marker of myofibroblast activation) or transforming growth factor-β (TGF-β, a pro-fibrotic marker) in either wild-type (Figures 5E,G) or Gabbr1 ^cKO mice (Figures 5F,H). Collectively, these findings suggest that acute GABA_BR/GLP-1R activation does not significantly modulate cardiac fibrosis progression after MI.

GABA_BR activation enhances potassium conductance through inwardly rectifying GIRK channels in a G protein-dependent manner, and baclofen is sufficient to induce this current. To assess whether GLP-1RAs modulate GABA_BR-mediated inward currents, we performed whole-cell patch-clamp recordings in hiPSC-derived cardiomyocytes. Baclofen evoked inward currents, which were suppressed by GLP-1R activation in a concentration-dependent manner (Figures 6A,B). In addition, re-analysis of single-cell RNA sequencing data from cardiac tissue of 46 arrhythmia patients (Matthew C. Hill et al.) revealed co-expression of GABABR and GLP-1R in cardiomyocytes (Figure 6C) (Marso et al., 2016). Together, these findings suggest that GLP-1RAs attenuate GABABR-gated depolarizing currents in cardiomyocytes.