FGF13 Regulates VGSC-Independent Cardiomyocyte Impulse Propagation via Cx43 Trafficking

Abstract

Background:

Fibroblast growth factor homologous factor (FHF) variants associate with arrhythmias. Although FHFs are best characterized as regulators of voltage-gated sodium channel (VGSC) gating, recent studies suggest broader, non-VGSC-related functions, including regulation of Cx43 gap junctions and/or hemichannels, mechanisms that have generally been understudied or disregarded.

Methods:

We assessed cardiac conduction and cardiomyocyte action potentials in mice with constitutive cardiac-specific Fgf13 ablation (cFgf13^KO) while targeting Cx43 gap junctions and hemichannels pharmacologically. We characterized FGF13 regulation of Cx43 abundance and subcellular distribution. With proximity labeling proteomics, we investigated novel candidate mechanisms underlying FGF13 regulation of Cx43.

Results:

FGF13 ablation prolonged the QRS and QT intervals. Carbenoxolone, a Cx43 gap junction uncoupler, markedly prolonged the QRS duration leading to conduction system block in cFgf13^KO but not in wildtype (WT) mice. Optical mapping revealed markedly decreased conduction velocity (CV) during ventricular pacing. Microscopy revealed perturbed trafficking of Cx43, reduced localization in the intercalated disc, and suggested decreased membrane Cx43 but increased Cx43 hemichannels in cardiomyocytes from cFgf13^KO mice. Resting membrane potential (RMP) was depolarized and APD50 was prolonged in cFgf13^KO cardiomyocytes. Both were restored towards WT values with Gap19 (a Cx43 hemichannel inhibitor), expression of FGF13, or expression of a mutant FGF13 incapable of binding to VGSCs, emphasizing VGSC-independent regulation by FGF13. To assess the functional impact of RMP depolarization, hearts were subjected to hypokalemia, which had no effect in WT hearts but fully rescued CV in cFgf13^KO hearts. Proteomic analyses revealed candidate roles for FGF13 in the regulation of vesicular-mediated transport. FGF13 ablation destabilized microtubules and reduced the expression of tubulins and MAP4, the major cardiac microtubule regulator.

Conclusions:

FGF13 regulates microtubule-dependent trafficking and targeting of Cx43 and impacts cardiac impulse propagation via VGSC-independent mechanisms.

Graphical Abstract

INTRODUCTION

The fibroblast growth factor homologous factors (FHFs) are a subset of the fibroblast growth factors (FGFs) that are functionally distinct from canonical FGFs. Lacking signal sequences, FHFs (FGF11-FGF14) remain intracellular, are not secreted, do not bind to FGF receptors, nor function as growth factors ¹. In the heart, FHFs have been best characterized as cytoplasmic interactors and regulators of voltage-gated sodium channels (VGSCs) and are linked to arrhythmogenesis ^2–8. Human cardiomyocytes express both FGF12 and FGF13, with a FGF12 predominance ². A variant in FGF12 that perturbs the interaction with the main cardiac VGSC Na_V1.5 has been associated with Brugada syndrome ⁶ and, conversely, a mutation in Na_V1.5 that reduces the channel’s binding affinity for FGF12 has been associated with another inherited arrhythmia syndrome ⁸. More recently, a genome-wide association study for atrial fibrillation identified FGF13 as a risk locus ⁹ and decreased left atrial expression of FGF13 was associated with postoperative atrial fibrillation ¹⁰. Whether these FHF-associated arrhythmias arise solely from voltage-gated sodium channel dysfunction, however, is unclear. Attempts to define the full complement of effects mediated by FHFs and interpret the consequences of FHF variants on cardiac physiology and disease have been limited.

Connexin 43 (Cx43), the main connexin in ventricular cardiomyocytes, forms gap junctions at the intercalated discs (IDs), where VGSCs concentrate, allowing cell-cell electrical communications ¹¹. In mice in which FGF13, the predominant cardiac FHF in rodents, was ablated globally, cardiac conduction showed increased sensitivity to gap junction blockers or reduced levels of Cx43 ¹², effects attributed to reduced sodium channel conductance unmasked by pharmacological inhibition or genetic downregulation of Cx43 as a secondary hit, yet emerging evidence suggests that FGF13 participate in additional cellular processes. ^13–16 Here, we hypothesize that FGF13 modulates cardiac conduction directly through Cx43-dependent mechanisms. Building upon data showing that FGF13 stabilizes brain microtubules ¹³, we investigated whether FGF13 regulates intracellular trafficking of Cx43 and identify this as a new role for FGF13 in ventricular cardiomyocytes with consequences on cardiac conduction, independent of VGSC regulation.

Methods:

Data availability

All data will be provided upon request.

FGF13 constitutive knockout animals

To generate cardiac-specific constitutive knockout mice (cFgf13^KO) of the X-linked Fgf13, we crossed female Fgf13^{fl/fl 16} with hemizygous male Myh6-Cre (alpha-Myh6-Cre; The Jackson Laboratory, #011038) mice to generate hemizygous male knockout animals. Mice were maintained on the C57BL/6 background. Experiments were performed on 6- to 12-week-old mice.

Generation of cardiac-specific TurboID-FGF13VY transgenic mice

The TurboID-hFGF13VY construct was generated by fusing the cDNA of TurboID ³², containing a Hemagglutinin (HA) tag and a V5 tag at the N-terminal domain of the hFGF13VY cDNA. The construct was then cloned into the modified murine α-myosin heavy chain (MHC) tetracycline-inducible promoter vector ^41,42 to generate TurboID-hFGF13VY transgenic mice with non-targeted insertions of the tetracycline-regulated transgene in a C57BL/6 background. These mice were crossed with cardiac-specific (α-MHC), doxycycline-regulated, reverse transcriptional trans-activator (rtTA) mice to generate double transgenic mice ⁴³.

Generation of cardiac-specific TurboID-SCN5A transgenic mice

The V5-TurboID-NaV1.5 knock-in model was generated by Cyagen. The Kozak-3X HA tag-TurboID-V5 tag (V5-TurboID) was created by gene synthesis and inserted upstream of the ATG start codon, which is in exon 2 of the Scn5a gene located on mouse chromosome 9. The targeting vector included homology arms generated by PCR using BAC clone RP23–386N9 or RP23–198L19 with V5-TurboID and a Neo cassette flanked by self-deletion anchor (SDA) sites. Following successful targeting, diphtheria toxin A was employed for negative selection. The targeting construct was electroporated into embryonic stem cells. Positive clones were identified by PCR and confirmed by Southern blotting. Five clones were injected in C57BL/6N blastocysts, and chimeras were crossed to C57BL/6 mice. Heterozygous mice were bred to produce homozygous offspring.

In vivo biotinylation

Male and female double transgenic mice, 4–8 months old, were placed on a doxycycline diet (in pelleted form, 200mg/kg of diet) for 10 days to express the TurboID-FGF13VY transgene in the heart. Biotin (10 μl/g from a 2.4 mg/ml stock (in PBS:DMSO 9:1) or 10% DMSO in PBS (for the controls)) were injected daily, subcutaneously, for 3 consecutive days to achieve biotinylation in the hearts of TurboID-FGF13VY and V5-TurboID-Na_V1.5. The mice were sacrificed to collect the hearts 24 hours after the third injection. Whole heart tissues were lysed with a handheld tip homogenizer in Lysis Buffer (50 mM Tris-HCl pH 7.5, 150mM NaCl, 10mM EDTA, 0.5% Sodium Deoxycholate (m/v), 1% Triton X-100 (v/v), 0.1% SDS (w/v)) supplemented with Roche cOmplete Protease Inhibitor Cocktail and 1 mM PMSF (PhenylMethylSulfonyl Fluoride). Biotinylation efficiency was confirmed by Western blot analysis with Streptavidin-HRP.

Sample preparation for mass spectrometry

Heart tissue samples for mass spectrometry analysis were prepared as previously described ⁴⁴ and adapted for this study. Whole heart lysates were centrifuged at 21,130 × g at 4 °C for 15 minutes. 1 mg was precipitated with ice-cold trichloroacetic acid (TCA, 55% in double-distilled water) on ice for 15 minutes and centrifuged at 21,130 × g at 4 °C for 15 minutes. The pellets were washed with −20 °C cold acetone, resuspended and centrifuged at 21,130 × g at 4 °C for 10 minutes. Acetone was removed and the pellets washed with acetone again three times to ensure complete TCA removal. After the last acetone wash the pellets are resuspended in urea buffer (8M urea, 100 mM sodium phosphate pH 8, 100 mM NH₄HCO₃, and 1% SDS (w/v), centrifuged at 21,130 × g at room temperature for 10 minutes, transferred to new microcentrifuge tubes and 10mM Tris(2-carboxyethyl)phosphine was added to reduce disulfide bonds. Twenty mM of freshly prepared iodoacetamide was added to alkylate free cysteines, vortexed immediately and incubated for 25 minutes in the dark. Alkylation was then quenched by adding 50mM of dithiothreitol (DTT) and water was added to dilute the samples down to 4M urea and 0.5% weight/volume (w/v) sodium dodecyl sulfate (SDS). Samples were added to 50 μl of streptavidin magnetic beads pre-washed in 4 M urea, 100 mM sodium phosphate pH 8, 0.5% w/v SDS and tubes were rotated overnight at 4 °C to for streptavidin pull-down. The beads were then transferred to a new microcentrifuge tube, washed 3 times with 4 M urea, 100 mM sodium phosphate pH 8, 0.5% SDS (w/v) three times with 4 M urea, 100 mM sodium phosphate pH 8 without SDS. The beads were transferred to new tubes for one last wash step. Before the final bead pull-down 10% of the resuspended beads were collected for streptavidin-HRP blotting.

Mass spectrometry analysis

Mass spectrometric analysis for the FGF13 TurboID dataset was performed at the Weill Cornell Medicine Proteomics and Metabolomics Core Facility. In brief, the protein on beads were reduced with DTT, alkylated with iodoacetamide, and digested overnight with trypsin at 37 °C. The digests were desalted by C18 Stage-tip columns. The digests were analyzed using a Thermo Fisher Scientific EASY-nLC 1200 coupled on-line to a Fusion Lumos mass spectrometer (Thermo Fisher Scientific). Buffer A (0.1% FA in water) and buffer B (0.1% FA in 80 % ACN) were used as mobile phases for gradient separation. A 75 μm × 15 cm chromatography column (ReproSil-Pur C18-AQ, 3 μm, Dr. Maisch GmbH, German) was packed in-house for peptide separation. Peptides were separated with a gradient of 5–40% buffer B over 30 min, 40%−100% B over 10 min at a flow rate of 400 nl/min. The Fusion Lumos mass spectrometer was operated in a data independent acquisition (DIA) mode. MS1 scans were collected in the Orbitrap mass analyzer from 350–1400 m/z at 120K resolutions. The instrument was set to select precursors in 45 × 14 m/z wide windows with 1 m/z overlap from 350–975 m/z for HCD fragmentation. The MS/MS scans were collected in the Orbitrap at 15K resolution. Data were searched against the mouse Uniprot database (downloaded on August 7, 2021) using DIA-NN v1.8 and filtered for 1% false discovery rate for both protein and peptide identifications. Statistical significance was determined by multiple t-tests corrected for multiple comparisons (Bonferroni). Proteomic analysis for data of FGF13-TurboID is displayed in Figure 6 and Supplemental Dataset 1. TMT quantified proteomic analysis for data of Na_V1.5-TurboID as displayed in Figure 6 and in Supplemental Dataset 2 was performed as previously described for APEX-fusion proteins following an SPS-MS3 mass spectrometry method ⁴⁵. For Gene Ontology (GO) term analysis, proteins with a Log₂FC > 2, P_adj < 0.05 were used and g:Profiler ⁴⁶ was used to generate GO term tables.

Electrocardiograms

Mice were anesthetized with tribromoethanol (Avertin) and single lead ECGs were recorded subcutaneously, while their body temperature was maintained with a heat lamp at 37 °C. Signals were amplified and recorded with an Octal Bio Amp amplifier connected to a Powerlab 16/30 DAQ system (ADInstructions). Heart rate, PR, QRS and QT intervals were analyzed using LabChart, version 8.0, software (ADInstruments). The ECG Analysis Module for LabChart, with the following specifications, was used for interval analysis: Typical QRS width 10 ms, R waves > 60 ms apart, and Pre-P baseline 10 ms. Fifty to 100 beats were selected and analyzed with 4 beat averaging. Automated outputs were manually checked for accuracy using the ECG Averaging View and edited as needed. Conduction block was determined when there was no QRS complex after a P wave. For ECG recordings with carbenoxolone, carbenoxolone disodium salt (C4790, Sigma-Aldrich) was dissolved in distilled water (90 mg/kg) and injected intraperitoneally. Surface ECGs were recorded 30 minutes after carbenoxolone administration for 15 min.

Electrophysiology Recordings

Action potentials were recorded in an extracellular solution containing (in mM): 140 NaCl, 5.4 KCl, 1.8 CaCl₂, 1 MgCl₂, 10 glucose, and 10 HEPES, with pH adjusted to 7.4 with NaOH. Patch pipettes made from borosilicate glass (2–2.5 MΩ resistance) were filled with an internal solution consisting of (in mM): 125 K-gluconate, 20 KCl, 5 NaCl, 1 MgCl₂, 5 MgATP, and 10 HEPES, adjusted to pH 7.2 with KOH. All chemicals were purchased from Sigma-Aldrich. To compare action potentials between WT and KO myocytes, the perforated patch technique was employed with 200 μg/ml amphotericin B added to the internal solution and action potentials were elicited by injecting a 200 pA current pulse for 20 ms. For all other action potential recordings, the whole-cell patch clamp technique was used. In the FGF13 rescue experiments, cells were held at a membrane potential of −90 mV, and action potentials were recorded at rheobase. In GAP19 treatment experiments, cells were held at their resting membrane potential, and action potentials were induced by injecting a 200 pA current pulse for 50 ms.

Optical Mapping

Mice (8–12 wk old) were injected with 0.3 ml of a 1:1 heparin:saline solution and euthanized. Hearts were rapidly excised, cannulated, and perfused with oxygenated Tyrode’s solution (in mM: 129 NaCl, 24 NaHCO₃, 4 KCl, 1 MgCl₂, 11.2 mM glucose, 1.8 CaCl₂, 1.2 KH₂PO₄) on a Langendorff apparatus. The perfusate contained Blebbistatin (10 μM, Sigma-Aldrich, #B0560) to suppress motion artifacts and 30 μl of 2.5 mM Di-4-ANEPPS (ThermoFisher #D1199) was injected via inline port. Hearts were submerged in a 37 ± 1°C bath of Tyrode’s with perfusion pressure maintained at ~70 ± 15 mm Hg by adjusting flow (2.0 ± 1 ml/min).

A pacing electrode was placed in the left ventricular apex. ECG and high-resolution optical signals of the anterior epicardial surface of the ventricles were acquired as previously described with a 6×6 mm² field of view ⁴⁷. Diastolic threshold was determined at a pacing rate of 7.14 Hz; experiments were run at 1.5× diastolic threshold.

The hearts were subjected to a rapid pacing protocol under normal conditions and then the perfusate switched to low potassium Tyrode’s (2 mM KCl). The hearts were perfused with low potassium Tyrode’s for 20 min before rapid pacing protocol was repeated.

Rapid Pacing Protocol:

Pacing rate progressively increased from 8 to 24Hz with two-second recordings taken at each step. If there was loss of capture, voltage was increased until capture regained, or maximum voltage was hit with no capture and the protocol was ended. After switching to new perfusate, the voltage was reset to 1.5x diastolic threshold before beginning the rapid pacing.

Optical Mapping Analysis:

Optical recordings were analyzed using the open-source software ElectroMap in MATLAB ⁴⁸. The ventricles were manually traced as the region of interest. Spatial and signal processing settings included a 3×3 pixel Gaussian filter (sigma=1.5), Savitzky-Goaly filter, and Top-Hat average filter (length of 100) for baseline correction. For each pacing rate, 10 consecutive beats were analyzed, and conduction velocity (CV) was calculated through ensemble averaging. Data were graphed in GraphPad Prism 10.

Cardiomyocyte Isolation, Cell Culture and Adenoviral Transduction

Adult murine cardiomyocytes were isolated from left and right ventricles of Myh6-Cre⁺ (WT) and Myh6-Cre+;Fgf13^fl/fl (cFGF13KO) male mice using a previously validated Langendorff-free cell isolation method ⁴⁹. Briefly, mice aged 8–12 weeks were anesthetized with Avertin and the chest was incised to expose the heart. The inferior vena cava and descending aorta were cut, and the heart was flushed by injecting 10 ml of room temperature EDTA buffer into the right ventricle. EDTA buffer contains (in mM): 130 NaCl, 5 KCl, 0.5 NaH₂PO4, 10 HEPES, 10 Glucose, 10 2,3-butanedione 2-monoxime (BDM), 10 Taurine, 5 EDTA. After placing an aortic clamp, the heart was excised and transferred to a 60-mm dish containing EDTA buffer. Next, 10 ml of room temperature EDTA buffer was injected into the apex region of the left ventricle. The same aperture was used to inject 5 ml of perfusion buffer (warmed to 37C) and then serial injection of collagenase buffer (5X 10 ml tubes with collagenase buffer) into the left ventricle (LV). Perfusion buffer contains (in mM): 130 NaCl, 5 KCl, 0.5 NaH₂PO₄, 10 HEPES, 10 Glucose, 10 BDM, 10 Taurine, 1 MgCl_2. The atria were then separated and discarded. The ventricles were pulled into 1-mm pieces with forceps and cells were dissociated with gentle trituration for 2 minutes. Enzymatic activity was inhibited by addition of 3 ml of stop buffer (10% FBS in Perfusion Buffer), and the cell suspension was passed through a 150-mm filter and allowed to gravity settle for 20 minutes. The cells were then washed twice in perfusion buffer and allowed to gravity settle for 10 minutes each and used for various experiments.

Cell culture and adenoviral infection of adult mouse ventricular myocytes were performed as described ^2,49. Briefly, prior to cardiomyocyte isolation, 12-well plates were coated with laminin at a concentration of 5 mg/ml in PBS, for 1 hour at 37°C. After isolated cardiomyocytes were retrieved from gravity settling, the laminin solution was aspirated, the wells were washed with PBS, and cardiomyocytes equilibrated in Plating Media (described in ref. ⁴⁹) were added to the wells. Cells were allowed to adhere for 1 hour. Subsequently, the Plating Media was removed, and the cells were incubated in 500 ml of Culture media per well. After 2 hours in Culture media described in ref. ⁴⁹), a subset of KO cardiomyocyte wells was infected with 1.5 μl of FGF13-VY adeno-associated virus serotype 8 (AAV8, 1.6 × 10⁸ ifu/ml) or 1.5 μl of FGF13^R/A AAV8 (6.6 × 10⁷ ifu/ml). The control group was infected with an AAV8 expressing GFP. Cardiomyocytes were incubated in Culture Media for 48 hours for adequate viral expression before subsequent analyses were performed. Electrophysiological analyses were obtained from recording sessions in which all relevant experimental groups had been simultaneously cultured. The amount of AAV-expressed FGF13 or FGF13^R/A in the cFgf13^KO cells was similar to the level of endogenous FGF13, as shown previously ²².

Immunoblots

Isolated cardiac myocytes were lysed for 30 minutes in ice cold RIPA buffer supplemented with Halt protease and phosphatase inhibitor cocktails (Thermo Fisher Scientific). The homogenate was then centrifuged at 17,000 g at 4C for 15 minutes. Supernatant was collected and protein concentration was determined by bicinchoninic acid assay. Protein was separated on 8–16% gradient Tris-Glycine gels (Thermo Fisher Scientific) for 100 minutes at 140V. Once separated, gels were transferred to PVDF membranes using iBlot 3 Western Blot Transfer System (Thermo Fisher Scientific) in Broad Range mode for 6 minutes. Membranes were immunoblotted for Na_V1.5 (1:1000, 493–511, Alomone Labs), FGF13 (1:1000, custom designed ²) GAPDH (1:1000, MA5–15738, Thermo Fisher Scientific), Cx43 (1:1000, C6219, Sigma-Aldrich), ZO-1 (1:500, 5406, Cell Signaling Technology), vinculin (1:1000, sc-73614, Santa Cruz Biotechnology), Na⁺/K⁺ ATPase (1:1000, sc-48345, Santa Cruz Biotechnology), α-tubulin (1:1000, sc-5286, Santa Cruz Biotechnology), β-tubulin (1:1000, 2146, Cell Signaling Technology), acetylated α-tubulin (1:1000, 5335S, Cell Signaling Technology), and MAP4 (1:500, 11229–1-AP, Proteintech). All primary antibodies were incubated overnight at 4 °C at respective concentrations. HRP-tagged secondary antibodies were incubated at room temperature for 1 hour. The blots were visualized by chemiluminescence, and images were captured using ChemiDoc Tough Imaging System (Bio-Rad). Twenty-thirty μg of protein was loaded per well when probing for most proteins. For Cx43, we used 5 μg of protein to better visualize the different bands associated with the different phosphorylation states.

Co-immunoprecipitation

Co-immunoprecipitation shown in Figure 2A was performed as following: hearts from 3 WT mice, 4–6 months old, were collected and lysed in NP40 Lysis buffer (150mM NaCl, 20mM Tris-HCl pH 8, 2mM EDTA, 1% NP40 (v/v)) supplemented with Roche cOmplete Protease Inhibitor Cocktail and 0.2 mM PMSF. One mg of lysate was pre-cleared with Dynabeads Protein A beads (Thermo Fisher Scientific) for 1 h at 4 ºC. Pre-cleared lysates were incubated with 5 μl of Nav1.5 antibody (Alomone #493–511) (0.8 μg/μl, Alomone Labs) overnight at 4 ºC. Immunoprecipitation was performed with 50 μl of Dynabeads Protein A beads for 1 h at room temperature and proteins were eluted with 50 μl of 0.2 M glycine at pH 2.4 for 15 min at room temperature, immediately neutralized with an equal volume of 1 M Tris-HCl, pH 8.

Co-immunoprecipitation for Figure 7 was performed according to the manufacturer’s instruction (Capturem IP & Co-IP Kit, Takara, 635721). Briefly, isolated ventricular myocytes (nearly 5X 10⁵ per genotype) were lysed on ice for 30 min with 200 μl lysis buffer. Following 17,000 × g centrifugation at 4C for 10 min, the supernatant was incubated with 10 μg antibody (maintaining an antibody: lysate concentration of 2:500) for 20 minutes at room temperature. After incubation, the antibody-bound lysate was added onto the spin column and centrifuged at 1000 × g for 1 min at room temperature. Then, 100 μl wash buffer was added to the spin column and centrifuged at 1000 × g for 1 min again. Finally, 50 μl elution buffer was added to the spin column and centrifuged at 1000 × g for 1 min. The eluted sample was used for Western blot analysis as previously described ³³.

Quantitative Real-Time PCR

RNA was extracted from isolated ventricular cardiomyocytes using instructions from the RNeasy Mini Kit (QIAGEN). Reverse transcription was performed using Bio-Rad’s instructions from the iScript cDNA Synthesis Kit. For each sample, qPCR was performed in duplicate using SYBR green-based detection (Bio-Rad). Ct values were quantified using Quantstudio 3 (Applied Biosystems). Gapdh or Vcl were used as a reference gene. PCR primer sequences for target genes are as follows:

	Primers for qRT-PCR
Gene	Forward primer (5’ to 3’)	Reverse Primer (3’ to 5’)
Gja1	GGTGATGAACAGTCTGCCTTTCG	GTGAGCCAAGTACAGGAGTGTG
Map4	CAGTCTTGTGGATGCGTTGAC	TTCCCGGTTTTCTCATCACCA
Tuba4a	CACCATTGGCAAGGAGATCATCG	CATCAGCAGAGAGGTGAAGCCA
Gapdh	TGTCAGCAATGCATCCTGCA	CCGTTCAGCTCTGGGATGAC
Vcl	CCTATCAAGCTGTTGGCAGTAGC	TGTGGCTCCAAGCCTTCCTGAA

Histological Analysis

Hearts were rinsed in PBS, fixed in 4% paraformaldehyde overnight at 4C, and dehydrated in a series of ethanol washes. Samples were cleared in xylene and mounted in paraffin. Sections of 10 μm were cut and stained with hematoxylin and eosin to assess gross morphology. Samples from different groups were processed in parallel, and histological analyses performed. Immunostaining and confocal microscopy were performed as described ⁵⁰. For each antibody, secondary antibody-only controls were performed at 2 different concentrations (1:500 and 1:1000) to confirm genuine target staining.

Echocardiography.

Echocardiography was performed on conscious mice by a technician blinded to animal genotype as previously described ¹⁵. In brief, images were obtained on a VEVO 2000 high-resolution imaging system (VisualSonics). Transthoracic 2D M-mode used for data analysis. Heart rate, left-ventricular end-diastolic diameter (LVEDD), and left-ventricular end-systolic diameter (LVESD) were measured from at least three consecutive cardiac cycles by two experimenters blinded to genotype. Fractional shortening (FS) was calculated with the formula: FS = (LVEDD – LVESD)/LVEDD.

Quantitative Immunocytochemistry Analysis

Isolated cardiomyocytes were fixed for 10 minutes in 4% paraformaldehyde and immunolabeled with antibodies against Cx43 (C6219, Sigma, St Louis, MO) and N-cadherin (sc-59987, Santa Cruz Biotechnology). Images were captured on a Zeiss LSM 800 confocal microscope and the amount of immunoreactive signal at the ICD was analyzed with NIH ImageJ/FIJI software as described previously ⁵¹. In addition, using a Pearson-Spearman correlation colocalization plug-in, quantitative statistical colocalization on the two-color confocal images was performed as described ⁵². For ZO-1-WGA colocalization image analysis, hearts from 2-month-old WT and FGF13 KO mice were harvested, rinsed in PBS, and incubated sequentially in 15% and 30% sucrose solutions. Hearts were then incubated in OCT medium (TissueTek) for 1 hour and then frozen in OCT. Ten micron cryosections were made, permeabilized with 0.1% Tween-20 + PBS (PBST), and blocked in PBST + 2% Fetal Bovine Serum (FPBST) for 1 hour. Primary antibodies were incubated overnight with respective secondary Alexa-Fluor antibodies for 1 hour at room temperature. After mounting, images were obtained at 20X using Thunder Leica Microscope. In ImageJ, the red (WGA) and green (ZO-1) channels were converted to 8-bit images. The ZO-1 mean pixel intensity per image was measured. Then the WGA intensity was subtracted from ZO-1 and the residual pixel intensity was measured. The residual intensity indicated the amount of non-colocalized ZO-1. After initial optimization and development of analytical algorithms, the experimenter was blinded to genotype for image analyses.

Biotinylation

Biotinylation was performed according to the manufacturer’s instruction (Pierce™ Cell Surface Biotinylation and Isolation Kit, ThermoFisher Scientific, A44390).

Area and Pixel Intensity Calculations

Isolated cardiomyocytes were fixed in 4% paraformaldehyde for 10 minutes and Cx43 was identified with anti-Cx43 antibody (C6219, Sigma, St Louis, MO). Images were captured in 40X using a Leica immunofluorescence microscope. The images were opened in NIH/ImageJ software, converted to 8-bit images, and a black and white threshold was applied. Then the cell border was manually drawn, and the area was measured as a percentage of the cell. For pixel intensity calculations of confocal images, a max intensity z-stack was generated. Then the background was subtracted. A line was drawn along the longest axis of the cell, and pixel distribution generated as a Plot Profile. For calculation of Cx43 puncta size from confocal images, a max intensity z-stack was generated. In ImageJ/FIJI, each image’s background was subtracted, converted to 8-bit, and thresholded. The lower threshold was set to 22 (based on clear visibility of small and large puncta with minimal pixel overlap) and applied to all images. The number of particles was counted using the “Analyze Particles” function. Small puncta were defined as between 0 – 0.2 μm² (based on literature showing that Cx43 loaded vesicles are <150 nm in diameter) and large puncta were counted between 0.21 – 1.80 μm².

Statistical Analysis

Statistical analyses were performed with Graphpad Prism (version 10.6.1). Results are presented as mean ± standard error. For datasets with n<6, non-parametric tests were applied. For datasets n≥6, parametric or non-parametric tests were applied after testing for normality. For all analyses with within-test multiple testing, multiple test correction was applied. The statistical significance of differences between groups was assessed using either a two-tailed Student t test or one-way analysis of variance and was set at P<0.05. Other statistical analyses were performed as previously described ¹⁵ or as described in corresponding figure legends. No animals were excluded from analyses. Representative images were selected to best illustrate the quantified average/mean in corresponding graphs.

RESULTS

Electrocardiograms and optical mapping of cardiac-specific Fgf13 knockout mice suggest voltage-gated sodium channel-independent regulation

To explore non-VGSC roles of the X-linked FGF13 in the heart, we generated constitutive, cardiac-specific Fgf13 knockout (cFgf13^KO) mice by crossing females from a validated Fgf13^fl/fl mouse line ¹⁶ with Myh6-Cre⁺ males. Efficient knockout was confirmed by western blot and quantitative RT-PCR (Supplemental Figure 1A-B). Loss of FGF13 was not associated with a significant change in heart weight/body weight or gross morphology (Supplemental Figure 1C-D). Echocardiograms on conscious mice showed no significant genotype differences in left ventricular function (Supplemental Figure 1E-F), however, ECGs showed that cFgf13^KO mice have longer QRS durations and QT intervals (Figure 1A-B). These results suggest that baseline electrical activation through the conduction system—and possibly the ventricular myocardium—is delayed by FGF13 knockout. This finding is consistent with observations in inducible Fgf13 knockout ¹⁶ and after Fgf13 knockdown ², although not in a global Fgf13 knockout model ¹². The reason(s) for the different effects across models remain unclear, but because several studies showed that global Fgf13 ablation is not viable^15,17–19, we suspect that the reported global knockout mice retain residual Fgf13 expression, potentially attenuating the phenotype.

Figure 1. Fgf13 knockout prolongs the QRS duration, increases sensitivity to gap junction blockade and slows conduction velocity.

A. Representative lead I ECGs at baseline. B. Summarized data (N = 11 mice each) for ECG parameters, 2-tailed, unpaired t-tests with Welch’s correction. C. Representative ECGs at baseline, and 30, 35, 40 minutes after 90 mg/kg i.p. carbenoxolone (CBX), N = 5 mice each). D. Summarized data (N = 5 each) for QRS duration at 30 minutes post-CBX injection. Multiple Kolmogorov-Smirnov tests corrected for multiple comparisons. E. Quantification of conduction block occurrence after CBX administration. All KO mice (5/5) developed conduction block with 90 mg/kg CBX. WT were unaffected. Fisher’s exact test. F. Representative isochrones demonstrating slower baseline CV in cFgf13^KO hearts. G. CV (N = 5 WT and 4 cFgf13^KO mice) at 8 Hz pacing rate, Mann-Whitney test. H. Slowed CV in the cFgf13^KO hearts is significant at all pacing rates tested until capture was lost, 2-way ANOVA corrected for multiple comparisons. I. Critical pacing rate, Mann-Whitney test.

While the QRS prolongation in cFgf13^KO mice could be from known effects on VGSC functions and/or availability attributed to FGF13 ^2,3,20, data showing that FGF13 affects voltage-gated potassium channels ¹⁶ and contributes to cellular functions beyond regulation of VGSCs ^13–16 prompted us to consider whether QRS prolongation reflects broader effects of FGF13. Specifically, since the Fgf13 global knockout mice showed increased sensitivity to pharmacological inhibition of Cx43 by the gap junction blocker carbenoxolone or Cx43 genetic knockdown in cardiomyocytes ¹², we focused on Cx43 as a primary FGF13 target, independent from VGSCs. To test this, we recorded surface ECGs after treatment with 90 mg/kg carbenoxolone, a dose at which ECG changes were detected in the whole-body knockout model. WT littermates were not appreciably affected by carbenoxolone, but all cFgf13^KO mice showed not only progressive QRS prolongation but also high-degree atrioventricular block (Figure 1C-E, Supplemental Figure 2).

While QRS duration primarily reflects activation through the specialized conduction system, myocardial conduction can be directly measured using optical mapping during ex vivo ventricular pacing with which we found that conduction velocity (CV) was significantly slower in cFgf13^KO ventricles compared to WT (Figure 1F–G). To further explore the mechanism, we assessed rate-dependent changes in conduction with varying pacing frequencies. At every rate tested, including the slowest (8 Hz), CV remained consistently lower in cFgf13^KO hearts (Figure 1G–H). Notably, the extent of CV slowing was comparable at both basal and rapid pacing rates, arguing against a role for impaired rate-dependent recovery from Na⁺ channel inactivation—a known consequences of FGF13 loss. Moreover, the critical pacing rate, the highest frequency at which 1:1 capture was maintained—was significantly reduced in cFgf13^KO hearts (Figure 1I), indicating impaired conduction reserve.

The previously reported whole-body knockout model showed no QRS widening nor atrioventricular block even at a higher dose of carbenoxolone (120 mg/kg), suggesting that those mice were less sensitive to carbenoxolone, so we hypothesized that they have a larger Cx43 reserve. Indeed, conduction block in those mice was only detected when they were also haploinsufficient for Cx43. The etiology of the larger Cx43 reserve is not known but is consistent with our hypothesis that those mice have residual FGF13 and consequent reduced effects on Cx43. Our results thus suggest that impaired Cx43-mediated conduction through gap junctions and/or Cx43 hemichannels contributes to the QRS widening, atrioventricular block, and slowed conduction velocity when Fgf13 is eliminated.

Fgf13 knockout hearts show altered Cx43 protein expression and spatial patterning

We examined whether FGF13 ablation impacts Cx43 binding to Na_V1.5. As they are both known direct Na_V1.5 interactors ^5,21, we reasoned that if the effect of FGF13 ablation on cardiac activation is due to perturbed VGSCs, then the Cx43-Na_V1.5 interaction would remain unaltered. Instead, in an analysis of a dataset cataloging proteins co-immunoprecipitated with Na_V1.5 in WT vs. cFgf13^KO hearts ²², we observed a 63% reduction (P=0.009) in Cx43 co-immunoprecipitated in cFgf13^KO hearts, suggesting that less Cx43 is in complex with Na_V1.5 in cFgf13^KO hearts and altered subcellular Cx43 localization.

We therefore systematically investigated the consequences of Fgf13 knockout on Cx43 protein in ventricular tissue and in isolated ventricular cardiomyocytes. As expected, Cx43 in WT hearts was highly enriched at IDs, marked by N-cadherin. In cFgf13^KO heart tissue, however, we detected Cx43 in large puncta that did not co-localize with N-cadherin (reduced Spearman’s correlation coefficient, Figure 2A-B). N-cadherin, in contrast, was not significantly affected by loss of FGF13 (Supplemental Figure 3A-D). Much of the Cx43 signal in cFgf13^KO sections localized to the cardiomyocyte lateral membranes (marked by wheat germ agglutinin [WGA]) rather than to the IDs (Figure 2C), indicating a likely increased abundance of Cx43 hemichannels—Cx43 subunits along the lateral membranes that are not docked to other Cx43 subunits from adjacent cells and thus do not form gap junctions ²³.

Figure 2. FGF13 ablation perturbs Cx43 cellular distribution.

A. Representative images from paraffin-embedded ventricular tissue sections immunostained for Cx43 (green) and N-cadherin (red). Scale bar, 10 μm. B Spearman’s colocalization coefficient between Cx43 and N-cadherin (N = 3 mice each, WT n = 15 sections, KO n = 21 sections), nested t-test. C. Representative images from paraffin-embedded ventricular tissue sections immunostained for Cx43 (green) and wheat germ agglutinin [WGA] (red). Scale bar, 10 μm. D. Representative confocal images (z-stack max projection) of isolated ventricular myocytes immunostained with Cx43 (green). Insets: magnified areas showing different populations of Cx43 puncta. Scale bar, 10 μm. E. Relative abundance of Cx43 secretory vesicles (0–0.2 mm²) and annular gap junctions (0.21–1.8 mm²) from confocal images of WT (gray) and KO (red) ventricular myocytes (N = 4 mice each, WT n = 25 cells, KO n = 26 cells), nested t-test. F. Representative orthogonal slices from confocal images of WT and KO isolated cardiomyocytes immunostained for Cx43 (green) and lateral membrane protein Dystrophin (Dmd, red). White arrows: non-lateral membrane bound Cx43 particle in a KO myocyte. Scale bar, 5 μm. G. Representative images of isolated ventricular myocytes immunostained for Cx43 (green) and N-cadherin (red). Scale bar, 20 μm. H. Spearman’s colocalization coefficient between Cx43 and N-cadherin from isolated myocytes (N = 3 mice each, WT n = 17 cells, KO n = 21 cells), nested t-test. I. Pixel intensity across longitudinal axes of cardiomyocytes (N = 4 mice each, WT n = 28 cells, KO n = 26 cells). Cell distance and intensity are normalized to WT. Scale bar, 12 μm. J. Cx43 pixel intensity along the cell quantified using Area Under the Curve (AUC): Whole-cell (black line), Mid cell (10 – 90% normalized distance units, blue line), Cell edge (0–5 and 95–100% normalized distance units, purple line), nested t-test.

With freshly isolated single ventricular cardiomyocytes, which offer a uniform orientation for accurate quantification, confocal microscopy revealed two populations of Cx43 punctae (Figure 2D-E) consistent with previous observations ²⁴: large punctae, like those observed in tissue and that display characteristics and morphology consistent with annular gap junctions after endocytosis; and small punctae that likely represent Cx43-loaded vesicles trafficking along microtubules. We quantified the relative abundance of likely annular gap junctions (defined as ≥0.5 μm diameter and average area of 1.6±0.3 μm²) and likely Cx43-loaded vesicles (defined as ≤150 nm in diameter) and found that both populations were larger in cFgf13^KO cardiomyocytes. Moreover, many of the Cx43 punctae in cFgf13^KO cardiomyocytes were within the cytoplasm and not associated with the sarcolemma, while Cx43 punctae in WT control cells were generally localized at the sarcolemma (Figure 2F and Supplemental Figure 4). We also observed many large Cx43 punctae scattered throughout cFgf13^KO isolated cells, while in WT cardiomyocytes Cx43 punctae predominantly colocalized with N-cadherin at the IDs (Figure 2G-H). There were also more Cx43 punctae along the lateral membrane in cFgf13^KO cardiomyocytes (likely hemichannels) than in WT cells. While acute cardiomyocyte isolation could perturb the localization of ID components, the consistent genotype-specific patterns and the abnormal Cx43 distribution in cFgf13^KO cells were similar to what we observed in tissue, validating the isolated cardiomyocyte results. We therefore exploited quantified the distribution of Cx43 along the long axis of isolated cardiomyocytes. Compared to WT cells, Cx43 signal intensity in cFgf13^KO ventricular cardiomyocytes was larger between the cell ends (i.e., between the IDs) (Figure 2I-J) and smaller at the cell ends.

Cx43 immunohistochemistry and immunocytochemistry in tissue and cells revealed both a genotype difference in Cx43 spatial patterning and suggested increased Cx43 abundance in cFgf13^KO myocytes. We therefore quantified the % area of the cardiomyocyte positive for Cx43 immunofluorescence (Figure 3A-B), which was greater in cFgf13^KO cardiomyocytes. Consistent with this, we found that Cx43 protein was 43 ± 9.4% higher in cFgf13^KO cells (Figure 3C-D). Although Cx43 total protein was increased, surface biotinylation performed with isolated cardiomyocytes showed that Cx43 was reduced at the cell surface by 28 ± 6% in cFgf13^KO cells (Figure 3E-F). This is consistent with our observation that Cx43 punctae in cFgf13^KO cardiomyocytes were generally within the cytoplasm and not at the sarcolemma (see Figure 2F and Supplemental Figure 4). Thus, we conclude that Fgf13 knockout increases Cx43 total protein yet decreases Cx43 abundance at the surface and perturbs Cx43 localization. To determine if the greater Cx43 protein abundance in cFgf13^KO cardiomyocytes resulted from increased Cx43 transcription, we performed qPCR and found no significant difference between WT and cFgf13^KO (Supplemental Figure 5A). We also analyzed Cx43 stability using cycloheximide applied to isolated cardiomyocytes to block translation. After three hours, we observed no significant difference in Cx43 degradation between WT and cFgf13^KO cells (Figure 3G-H), suggesting alternative mechanisms.

Figure 3. FGF13 knockout alters Cx43 protein expression.

A. Representative images of isolated cardiomyocytes immunostained with Cx43 (green) and respective 8-bit images with thresholding. Scale bar, 10 μm. B. Quantification of Cx43 expression from thresholded images approximated as % Cellular Area (N = 4 mice each, WT n = 30 cells, KO n = 35 cells), nested t-test. C. Immunoblot of isolated ventricular cardiomyocyte lysates (N = 3 mice each) immunoblotted for Cx43 and Vinculin. D. Immunoblot densitometric analysis of Cx43 protein expression normalized to vinculin from isolated ventricular cardiomyocytes (N = 10 mice each), 2-tailed, unpaired t-test with Welch’s correction. E. Immunoblot of Cx43 and Na⁺/K⁺-ATPase (loading control) biotinylated fractions (N = 4 mice each). F. Densitometric analysis of biotinylated Cx43 normalized to biotinylated Na⁺/K⁺-ATPase (N = 4 each), Mann-Whitney test. G. Representative immunoblot of untreated isolated ventricular cardiomyocyte lysates (0h) and 3-hour cycloheximide-treated (3h) immunoblotted for Cx43 and GAPDH. H. Densitometric analysis for Cx43 and GAPDH (N = 4 mice each). Mann Whitney test.

Action potentials in isolated cardiomyocytes suggests Fgf13 knockout increases Cx43 hemichannels

Recent data show that disease states like heart failure lead to an increase in Cx43 hemichannels and that the altered hemichannel distribution and activity affects cardiomyocyte excitability and contributes to arrhythmias ²⁵. Additionally, dystrophic mouse cardiomyocytes show that mislocalized and remodeled Cx43 hemichannels contribute to a depolarized resting membrane potential ^25–27 that can be hyperpolarized upon addition of the Cx43 hemichannel blocker, Gap19 ^26,27. Furthermore, pathological remodeling of Cx43 hemichannels can prolong the action potential duration, which can be reduced to nearly WT levels with Gap19 ²⁷. Overall, these studies implicate a critical role of Cx43 hemichannels in electrophysiological properties of cardiomyocytes.

Because we observed lateralization of Cx43 in tissue and cardiomyocytes from cFgf13^KO hearts, we tested for the consequences in cFgf13^KO mice by recording action potentials (APs) in isolated ventricular cardiomyocytes with the perforated patch technique. Consistent with data from an inducible knockout model, AP amplitude in cFgf13^KO myocytes was significantly reduced while rise time and APD50 (Figure 4A-D) were longer in cFgf13^KO myocytes compared to WT, likely due at least in part to a hyperpolarizing shift in the steady state inactivation V_1/2 of VGSC currents as previously observed ^2,16,28–30.

Figure 4. Fgf13 knockout increases aberrant Cx43 hemichannel contributions to action potentials.

A. Representative AP traces of 24–48 hour cultured ventricular cardiomyocytes using perforated patch clamp protocol from: WT (black), cFgf13^KO (red), cFgf13^KO + WT FGF13 AAV-mediated re-expression (blue), and cFgf13^KO + FGF13^R/A AAV-mediated re-expression (purple). Scale bar: x-axis 10 ms, y-axis 20 mV. B. AP Amplitude (WT N = 3 mice, n = 9 cells, KO N = 3 mice, n = 9 cells, KO + WT N = 3 mice, n = 8 cells, KO + FGF13^R/A N = 3 mice, n = 8 cells), nested 1-way ANOVA. C. AP rise time (WT N = 3 mice, n = 9 cells, KO N = 3 mice, n = 9 cells, KO + WT FGF13 N = 3 mice, n = 8 cells, KO + FGF13^R/A N = 3 mice, n = 8 cells), nested 1-way ANOVA. D. APD50 (WT N = 3 mice, n = 9 cells, KO N = 3 mice, n = 9 cells, KO + WT FGF13 N = 3 mice, n = 8 cells, KO + FGF13^R/A N = 3 mice, n = 8 cells), nested 1-way ANOVA. E. Resting membrane potential (RMP, WT N = 3 mice, n = 9 cells, KO N = 3 mice, n = 9 cells, KO + WT FGF13 N = 3 mice, n = 8 cells, KO + FGF13^R/A N = 3 mice, n = 8 cells), nested 1-way ANOVA. F. Representative APs (whole-cell patch clamp) of isolated myocytes injected with a 200-pA current pulse for 50 milliseconds. WT (black), WT + GAP19 (gray), KO (red), and KO + GAP19 (pink). Scale bar: x-axis 50 ms, y-axis 20 mV G. Summarized data (N = 3 mice each) for RMP in control (CTL) and GAP19-treated isolated cardiomyocytes (WT n = 19 cells, WT + GAP19 n = 15 cells, KO n = 21 cells, KO + GAP19 n = 17 cells), nested t-tests, 1-way ANOVA. H. APD50 (N = 3 mice each, WT n = 16 cells, WT + GAP19 n = 14 cells, KO n = 18 cells, KO + GAP19 n = 10 cells), nested t-tests, 1-way ANOVA. I. Peak amplitude (N = 4 mice each, WT n = 16 cells, WT + GAP19 n = 14 cells, KO n = 18 cells, KO + GAP19 n = 10 cells), nested t-tests, 1-way ANOVA. J. Representative images of ventricular myocytes cultured for 24–48 hours and immunostained with Cx43 (green). Scale bar, 10 μm. K. Cx43 pixel intensity analysis and normalized AUC (N = 3 mice each, WT n = 28 cells, KO n = 41 cells, KO + WT FGF13 n = 23 cells, KO + FGF13^R/A n = 16 cells), nested 1-way ANOVA.

We also found a more depolarized resting membrane potential (RMP) in cardiomyocytes isolated from cFgf13^KO cardiomyocytes, a parameter unlikely to be affected by VGSCs. To test that hypothesis definitively, we exploited a strategy to dissociate the contribution of FGF13 regulation of VGSCs by replacing WT FGF13 with a mutant version (Arg120Ala, “FGF13^R/A”) unable to bind to Na_V1.5 ^22,31. We infected cardiomyocytes isolated from cFgf13^KO mice with adeno-associated viruses (AAV) expressing either WT FGF13, FGF13^R/A, or GFP as a control. After 24–48 hours in culture to allow viral expression (all groups above were cultured in parallel, allowing direct comparisons), we observed that addition of either FGF13 or FGF13^R/A hyperpolarized the RMP to values similar to WT (Figure 4E).

We therefore hypothesized that the depolarized RMP in cFgf13^KO cardiomyocytes derived at least partly from increased Cx43 hemichannels, which we tested by adding the Cx43 hemichannel blocker, Gap19, which requires intracellular application and the whole-cell patch clamp technique to dialyze Gap19 via the patch pipette. The addition of Gap19 to the internal pipette solution hyperpolarized the resting membrane potential of cFgf13^KO cells but did not affect WT cardiomyocytes (Figure 4F-G), supporting our hypothesis.

Since excess Cx43 hemichannels contributed to the depolarized RMP in cFgf13^KO cardiomyocytes, we asked if these hemichannels affected other AP characteristics. We calculated the APD50, which showed an increased duration in cFgf13^KO cells (Figure 4H) and provide additional support for our hypothesis that excess hemichannels in cFgf13^KO cardiomyocytes affect AP properties and are consistent with a recent study showing that Cx43 hemichannels contribute to AP repolarization in isolated cardiomyocytes, especially in disease states in which hemichannels are more abundant ²⁵. Further, we examined the AP peak amplitude, a measure that depends mainly on Na_V1.5 channels (rather than Cx43 gap junctions). While in WT cells the peak amplitude was higher than in Fgf13^KO cardiomyocytes, addition of Gap19 to either WT or cFgf13^KO cells had no significant effect on the AP amplitude (Figure 4I), suggesting that Gap19 does not significantly affect parameters regulated mainly by VGSCs. Moreover, we observed that expression of either FGF13 or FGF13^R/A in cFgf13^KO cardiomyocytes ameliorated the observed abnormal Cx43 distribution, rendering Cx43 mainly localized to cell edges while markedly reducing scattered punctae throughout the cytoplasm (Figure 4J-K).

We next hypothesized that the depolarized RMP resulting from increased Cx43 hemichannel activity in cFgf13^KO hearts impairs CV by reducing Na_V1.5 channel availability. To test this, we repeated CV measurements with low extracellular potassium (2 mM), which hyperpolarize the RMP by approximately 20 mV. If conduction slowing under physiological K⁺ (4 mM) in the cFgf13^KO hearts is due to RMP-mediated VGSC inactivation, then hyperpolarization should restore VGSC availability and normalize CV. Indeed, under low K⁺ conditions, CV in cFgf13^KO hearts significantly increased across all pacing rates and became indistinguishable from WT hearts (Figure 5A–B). Similarly, the critical pacing cycle rate was normalized and matched values observed in WT controls (Figure 5C). These findings support the conclusion that RMP depolarization, driven by excess Cx43 hemichannel activity, reduces VGSC availability and contributes to slowed conduction in cFgf13^KO hearts. The restoration of CV under hyperpolarized conditions reinforces the functional link between RMP and myocardial conduction defects in this model and contrasts sharply with models in which slowed conduction arises from impaired recovery of Na⁺ channels from inactivation, conditions under which hypokalemia typically worsens, rather than improves, conduction, especially during rapid pacing.

Figure 5. Hyperpolarization of RMP by low K⁺ restores CV in cFgf13^KO hearts.

A. Representative isochrones showing baseline CV (WT N = 5 mice, cFgf13^KO N = 4 mice) with 2 mM extracellular K⁺. B. CV in the cFgf13^KO hearts is accelerated in 2 mM K⁺, not different from CV in WT hearts at all pacing rates tested. Data from normal K⁺ (4 mM) repeated from Figure 1 for comparison. C. Critical pacing rate at 2 mM extracellular K⁺, Mann-Whitney test.

That the Na_V1.5 binding incompetent FGF13^R/A restored Cx43 co-localization also suggested that at least some FGF13 functions are independent of its interactions with Na_V1.5. To test our hypothesis, we queried the relative Na_V1.5-free fraction of FGF13 in cardiomyocytes. Immunoprecipitation of Na_V1.5 successfully depleted 80% of the channel from cardiomyocyte lysates but only about 35% of FGF13 (Figure 6A-B), suggesting a stoichiometric excess of FGF13 compared to Na_V1.5. Thus, a significant fraction of cardiomyocyte FGF13 does not interact with Na_V1.5 and likely has other roles in cardiomyocytes.

Figure 6. FGF13 protein neighbors are associated with non-VGSC binding functions such as vesicle-mediated transport.

A. Immunoblot of WT heart lysates (N = 3 mice), Na_V1.5 coimmunoprecipitated fractions, and corresponding flowthrough (FT) fractions probed for Na_v1.5, FGF13, and GAPDH. B. Quantification of Nav1.5 CoIP shows that only about 35% of the total FGF13 is in complex, Mann-Whitney test, FT = Flowthrough. C. Left and middle panels are magnified images of left ventricular cryosections immunostained for FGF13 (red), N-cadherin (green) and WGA (green). Scale bar, 10 μm. Right, isolated myocytes with FGF13 TurboID and control myocytes immunostained for hemagglutinin (HA). Scale bar, 20 μm. D. Volcano plot of GO enriched terms (P_adj < 0.05, Log₂FC > 1). SCN5A is highlighted in a blue circle. Other proteins previously shown to coimmunoprecipitate with FGF13 are also indicated. Cx43 (Gja1) is also a significantly enriched FGF13 neighbor. E. Top GO terms for Biological Process. x axis, -log adjusted p value. F. Top KEGG terms. G. Venn diagram of FGF13 TurboID terms and SCN5A TurboID terms (Log₂FC > 2.0, P < 0.05). H. Top Biological Process GO terms from G separated into: FGF13 only no Na_v1.5 (salmon), Na_v1.5 only no FGF13 (purple), FGF13 and Na_v1.5 intersection (blue).

FGF13 protein neighbors suggest a role in vesicle-mediated transport of Cx43

To investigate how FGF13 affects Cx43 localization within cardiomyocytes and targeting to the IDs, we performed an unbiased screen for candidate FGF13 near neighbors. We used protein proximity labeling in whole hearts from transgenic mice expressing FGF13 fused to the biotin ligase TurboID ³² and a hemagglutinin (HA) tag. We first validated that the subcellular localization of transgenic TurboID-FGF13 mimics endogenous FGF13 by visualizing the TurboID-FGF13 via HA immunofluorescence in isolated ventricular myocytes. Like endogenous FGF13 immunostaining in control tissue sections (Figure 6C), HA immunofluorescence showed enhanced signal along the lateral membrane and at the IDs (marked by N-cadherin). Biotinylation efficiency of the TurboID-FGF13 in mice injected with biotin was confirmed by western blot with streptavidin-HRP, which showed intense signals over a large range of molecular weights compared to control mice injected with 10% DMSO (N = 3 mice, each), for which only endogenously biotinylated carboxylases were detected (Supplemental Figure 6A). Biotin-labeled proteins were captured by streptavidin and analyzed by semi-quantitative mass spectrometry. Compared to control samples without biotin, we identified 1549 unique proteins enriched in the FGF13 neighborhood (Log₂ fold change [Log₂FC]>2.0 and -log₁₀P_adj>1.30 [P_adj<0.05], Supplemental Dataset 1). As expected, we identified Na_V1.5 (Scn5a) (Figure 6D) and a set of proteins previously shown to coimmunoprecipitate with FGF13 from heart lysates ^15,33 such as junctophilin-2 (Jph2), cavin-2 (Cavin2), and cavin-4 (Cavin4). We also identified Cx43 (Gja1), further supporting for our hypothesis that FGF13 regulates Cx4 trafficking and targeting.

We performed gene set enrichment analysis on the entire set and ranked the gene ontology (GO) terms by their adjusted P-value, which revealed that among the GO biological processes (Figure 6E) many of the top terms were related to “cellular localization” (GO:0051641) and “vesicle-mediated transport” (GO:0016192). We also performed a gene enrichment analysis using the Kyoto Encyclopedia of Genes and Genomes (KEGG) (Figure 6F) and found that the top term was “endocytosis” (KEGG:04144). Together, these implicate FGF13’s role in bidirectional (anterograde and retrograde) transport processes.

Since an overarching goal was to identify FGF13 functions distinct from Na_V1.5 regulation, we performed a separate protein proximity labeling experiment with TurboID fused to Na_v1.5, allowing us to focus on the non-overlapping set of unique FGF13 near neighbors (Supplemental Dataset 2). The Na_V1.5 TurboID experiment, performed using tandem mass tags (TMT), identified 1015 proteins (Log₂FC>2.0, P<0.05), of which 642 were also contained in the FGF13 dataset, in which label-free quantification was applied. Among the FGF13 near neighbors, 907 proteins (58.5%) were absent in the Na_v1.5 dataset (Figure 6G-H), on which we then performed gene set enrichment analysis. Among the top hits were “intracellular transport” (GO:0046907), “vesicle-mediated transport” (GO:0016192), and multiple terms associated with microtubules (Supplemental Figure 7A). Gene set enrichment analysis of the 373 proteins unique to the Na_V1.5 dataset did not return any of these or related terms (Figure 6G-H). Thus, we focused additional studies on querying whether FGF13 regulates Cx43 via mechanisms related to vesicular transport.

FGF13-mediated microtubule instability alters Cx43 localization pattern

Building on these analyses and on previous data suggesting that FGF13 can stabilize microtubules ^13,34, we queried whether FGF13 affected microtubules as a mechanism to regulate Cx43 by first exploring an interaction between FGF13 and microtubules. In lysates from WT cardiomyocytes, alpha tubulin (but not beta tubulin) co-immunoprecipitated with FGF13 (Figure 7A), consistent with a recent study ³⁴. We then performed immunohistochemistry to evaluate alpha tubulin in tissue sections and found that the density and intensity of alpha tubulin differed between control and KO hearts (Figure 7B, Supplemental Figure 7B). Given the constraints in achieving consistent tissue orientation for quantitative analyses, we exploited acutely isolated ventricular cardiomyocytes. Using a thresholding protocol, we found that cFgf13^KO cells had lower tubulin density (Figure 7C-D). Further, we observed an increased immunofluorescence intensity of acetylated alpha tubulin, a marker of microtubule stability ³⁵, in cardiomyocytes isolated from WT (Figure 7E-F). Infection of cFgf13^KO cardiomyocytes with an AAV expressing FGF13 increased the acetylated tubulin immunofluorescence intensity, suggesting that re-expression of FGF13 plays a dynamic role in tubulin acetylation and microtubule stability (Figure 7E-F). Infection of cFgf13^KO cardiomyocytes with the Na_V1.5 binding-incompetent mutant FGF13^R/A also increased the acetylated tubulin immunofluorescence intensity, emphasizing that the effect is independent of FGF13’s interaction with Na_V1.5 (Figure 7E-F). We also quantified the relative fraction of acetylated alpha tubulin and total alpha tubulin by western blot. We observed a significant decrease in both in cFgf13^KO cardiomyocytes compared to WT cells (Figure 7G-I). Beta tubulin protein was similarly reduced in cFgf13^KO heart lysates compared to WT (Figure 7J). Immunofluorescence also showed decreased signal for beta tubulin in isolated cFgf13^KO cardiomyocytes (Figure 7K-L). Further, qRT-PCR revealed that transcripts for Tuba4a, one of the most highly expressed isotypes in the heart ³⁶, were significantly decreased (Supplemental Figure 5B). Thus, FGF13 knockout not only decreases microtubule stability but also reduces tubulin protein, suggesting that FGF13 exerts multiple regulatory controls on tubulins and microtubules.

Figure 7. FGF13 knokcout reduces microtubule stability and tubulin expression.

A. Coimmunoprecipitation (Co-IP) of FGF13 shows α-tubulin. Na_v1.5 is a positive control. Vinculin (VCL) for loading control. Protein concentration of IP fraction was 1/8^th of whole cell lysate (WCL). B. Representative 3D renderings of α-tubulin immunostain in ventricular tissue sections. Scale bar, 10 μm. C. Representative images of tubulin-stained myocytes, converted to 8-bit and thresholded. Scale bar, 15 μm. D. Summarized data of % cellular area of tubulin (n = 4 mice each, WT N = 47 cells, KO N = 47 cells), nested t-test. E. Representative images of myocytes immunostained with acetylated α-tubulin. Scale bar, 10 μm. F. Summarized data of acetylated α-tubulin (ACE) mean fluorescence (pixel) intensity (N = 4 mice each, WT n = 80 cells, KO n = 78 cells, KO + WT FGF13 n = 83 cells, KO + FGF13^R/A n = 51 cells), nested 1-way ANOVA. G. Representative immunoblots of ACE, α-tubulin, β-tubulin, and VCL (N = 3 mice each). H. Relative expression densitometric data of ACE normalized to α-tubulin from immunoblots (WT N = 10 mice, KO N = 8 mice), Mann-Whitney test. I. Densitometric quantification of α-tubulin normalized to VCL (WT N = 10 mice, KO N = 8 mice), Mann-Whitney test. J. Densitometric quantification of β-tubulin normalized to VCL (WT N = 6 mice, KO N = 5 mice), Mann-Whitney test. K. Representative confocal images of acutely isolated myocytes immunostained with β-tubulin. Scale bar, 15 μm. L. β-tubulin mean fluorescence (pixel) intensity (N = 3 mice each, WT n = 32 cells, KO n = 42 cells), nested t-test.

The overall regulation of tubulins and microtubules by FGF13 in cardiomyocytes prompted us to investigate the role of MAP4, a critical modulator of microtubule stability in cardiomyocytes, a key microtubule-associated protein implicated in cardiac health and disease ^37,38 and a protein identified in the TurboID screen as a candidate FGF13 near neighbor (Supplemental Figure 7C). We therefore assessed MAP4 in hearts (Figure 8A-B) and in isolated cardiomyocytes (Figure 8C-D) and found that MAP4 abundance trended towards a reduction in the knockout hearts. Additionally, Map4 mRNA was reduced in cFgf13^KO hearts (Supplemental Figure 5C). Overall, these results suggest that FGF13-deficiency in ventricular myocytes reduces the abundance of MAP4 and tubulin as well as markers of microtubule stability and perturbs trafficking of Cx43.

Figure 8. FGF13 knockout decreases MAP4 expression as well as impacts trafficking of ZO-1.

A. Immunoblot of MAP4 and GAPDH whole heart lysates (N = 3 mice each). B. Densitometric analysis of whole heart MAP4 protein expression (from A) normalized to GAPDH (N = 3 mice each), Mann-Whitney test. C. Immunoblot of MAP4 and Vinculin from isolated ventricular myocyte lysates (N = 3 mice each). D. Densitometric analysis of MAP4 protein expression from isolated myocyte lysates normalized to vinculin (WT, N = 5 mice; KO, N = 6 mice), Mann-Whitney test. E. Representative images of ventricular cryosections immunostained for ZO-1 (green) and WGA (red). Scale bar, 20 μm. F. Ccolocalization between ZO-1 and WGA in tissue sections (N = 3 mice each, 2–3 sections per mouse (WT n = 8 fields, KO n = 10 fields), nested t-test. G. Western blot of ZO-1 (vinculin as a loading control). H. Densitometric analysis of ZO-1 protein expression from isolated myocyte lysates normalized to vinculin (N = 4 mice each), Mann-Whitney test.

Since multiple proteins rely on vesicular transport via microtubules to reach their specific subcellular destinations, we hypothesized that FGF13 knockout affected the trafficking and targeting of proteins other than Cx43. In the FGF13 TurboID screen (Figure 4 and Supplemental Dataset 1), we identified multiple sarcolemmal ion channels and other ID proteins (Supplemental Table 1). Among these was the tight junction protein ZO-1 (Tjp1), a component of the ID and regulator of Cx43 organization ³⁹. We performed immunocytochemistry for ZO-1 and found that, like Cx43, ZO-1 distribution was perturbed (Figure 8E-F). However, unlike Cx43, ZO-1 expression was not significantly different between cFgf13^KO and WT hearts (Figure 8G-H).

DISCUSSION

Although FGF13 is best characterized as a regulator of VGSCs through FGF13’s direct interaction with their C-termini ^5,7, several reports suggest that FGF13 has additional. Among these, we previously determined that FGF13 regulates the abundance of K⁺ channels at the sarcolemma ¹⁶. With results here showing that FGF13 also directs the trafficking and targeting of Cx43 and ZO-1, we suggest that FGF13 is a regulator of trafficking and targeting of multiple proteins utilizing microtubule-dependent vesicular transport. Intriguingly, we found no difference in Gja1 transcription or Cx43 degradation despite the overall increased Cx43 abundance in cFgf13^KO cardiomyocytes. These results suggest that FGF13 may affect the rate or efficiency of Gja1 translation. Indeed, gene ontology for the TurboID dataset of terms found in the FGF13 near neighborhood but not in the Na_V1.5 near neighborhood identifies “regulation of translation” among the top Biological Process terms (GO:0006417; Supplemental Table 2).

While previous studies postulated that FGF13 affects Na_V1.5 trafficking and targeting via FGF13’s direct interaction with Na_V1.5, results here suggest that FGF13’s trafficking and targeting functions are more general. Indeed, we recently exploited the mutant FGF13^R/A, which is unable to bind to Na_V1.5, and showed that FGF13^R/A restored critical aspects of VGSC localization in cFgf13^KO cardiomyocytes via regulation of local membrane cholesterol ²². Thus, even some components of VGSC regulation by FGF13 are independent of FGF13 binding to the VGSC C-terminus. The ability of FGF13^R/A to restore trafficking and targeting of Cx43 suggests that FGF13’s effects on Cx43 are also independent of regulation of VGSC by FGF13.

The mechanisms by which FGF13 affects protein trafficking and targeting and thus increase the susceptibility of cFgf13^KO mice to arrhythmias in the setting of Cx43 blockade rely on microtubules and may include several independent regulatory steps. Specifically, we showed that FGF13 is in complex with, and regulates, microtubules in cardiomyocytes as well as controls the abundance of tubulins and the key cardiomyocyte microtubule regulator MAP4. Whether the reduction in MAP4 is upstream of microtubule instability or the reduction in tubulins and acetylated tubulins drives the decrease in MAP4 is unknown. We note that the microtubule deacetylases (HDAC6, SIRT2) were enriched in the FGF13 TurboID screen, and thus FGF13-dependent regulation of their activity may also contribute to the observed decrease in alpha tubulin acetylation in cFgf13^KO hearts and consequent microtubule stability. Our results suggest that the microtubule instability in cFgf13^KO hearts affects trafficking and targeting of Cx43 and other components of the ID. This leads to reduced Cx43 at the ID and increased Cx43 hemichannels at the lateral membrane, providing the substrate for arrhythmogenesis in the cFgf13^KO mice.

Our results showing that Cx43 subcellular localization is perturbed in cFgf13^KO cardiomyocytes and that increased Cx43 hemichannels contribute to slowed CV appear to contrast with a previous study that reported no difference in the localization of Cx43 in Fgf13 knockout myocytes ¹². That whole-body knockout model did report that Fgf13 knockout hearts were more sensitive to the gap junction inhibitor carbenoxolone, consistent with reduced Cx43 reserve. The reasons for the study differences are unclear, but may reflect different levels of Fgf13 reduction. Others ¹⁷ and we ^18,19 observed that complete Fgf13 knockout animals are nonviable ^15,17, in contrast to that Fgf13 whole body knockout study ¹², suggesting it may be a hypomorph with residual FGF13 expression. Consistent with the modeling data from that study, we also conclude that the failed impulse propagation with Fgf13 deficiency does not result only from a reduction in VGSC conductance. Results in Figure 1H show slowed CV at all pacing rates tested, even at slower rates at which VGSC inactivation is minimized, thus implicating mechanisms beyond reduced VGSC conductance ⁴⁰. At least one contributor likely is the depolarized RMP in cFgf13^KO cardiomyocytes, driving more VGSC inactivation due to the relatively steep steady-state inactivation curve near RMP ²². Further, our results suggest that underlying the depolarized RMP with Fgf13 deficiency is an increase in Cx43 hemichannels as shown by the restoration of RMP to WT levels with Gap19 (Figure 4G). Cx43 hemichannels were recently recognized as supporting inward depolarizing currents in disease states such as heart failure ^23,25, suggesting investigations into whether perturbation of FGF12, the homologous FHF predominant in human heart, contributes to arrhythmogenesis in heart failure.

In summary, our results suggest effects of FGF13 beyond regulation of VGSCs in cardiomyocytes and imply that multiple channels and auxiliary subunits important for normal rhythm are perturbed in Fgf13 knockouts. These results add to recent observations about the non-VGSC role for FGF13 in regulation of neuronal excitability ¹⁹, and emphasize the importance of exploring additional mechanisms of how FHFs contribute to cardiac physiology and how variants in FHFs lead to pathophysiology.

Novelty and Significance.

What Is Known?

• Fibroblast growth factor homologous factors (FHFs) bind to and regulate voltage-gated sodium channels (VGSCs) but have been implicated in additional cellular processes.

• Ablation of FGF13, the major FHF in rodent heart, increases sensitivity to Cx43 gap junction blockers and reduces cardiac conduction, which has been attributed to effects on VGSCs.

What New Information Does This Article Contribute?

• Fgf13 knockout in the mouse heart perturbs Cx43 gap junction trafficking and targeting and increases the abundance of Cx43 hemichannels; rescue with FGF13 or with an FGF13 mutant unable to bind to VGSCs restores Cx43 gap junction trafficking and targeting.

• Optical mapping under conditions of normal and low extracellular potassium shows that Fgf13 ablation slows cardiac conduction independent of effects on VGSCs.

• Proximity labeling proteomics reveals candidate non-VGSC near-neighbors and candidate mechanisms by which FGF13 affects Cx43 trafficking and targeting, including by stabilizing microtubules.

FHFs, members of the fibroblast growth factor (FGF) superfamily, have been extensively studied as direct interactors with, and regulators of, VGSCs and their effects (such as the association of FHF variants with arrhythmias) attributed to channel function even though other FGFs have not been implicated as ion channel regulators. Recent evidence implicates broader roles for FHFs, challenging the paradigm that arrhythmogenesis associated with FHF variants derives solely from VGSC dysfunction. This study shows that the primary cardiac FHF in mouse, FGF13, also regulates trafficking and targeting of Cx43 connexons and other components of the cardiomyocyte intercalated disc by controlling the stability of microtubules. Electrophysiological perturbations in cardiac-specific Fgf13 knockout cardiomyocytes are attributable, at least in part, to effects upon Cx43 gap junctions and an increase in Cx43 hemichannels. These observations not only highlight additional roles for FHFs, thus prompting further scrutiny of FHF functions in the heart, but also provide insight into the VGSC-independent mechanisms of arrhythmogenesis associated with FHF variants.

Nonstandard Abbreviations and Acronyms

AAV

adeno-associated virus

ACE

acetylated α-tubulin

AP

action potential

APD50

action potential duration at 50% repolarization

AUC

area under the curve

CBX

carbenoxolone

cFgf13^KO

cardiac-specific Fgf13 knockout

Co-IP

coimmunoprecipitation

CTL

control

CV

conduction velocity

Cx43

Connexin 43

FGF13^R/A

FGF13 Arg120Ala

FGFs

fibroblast growth factors

FHF

Fibroblast growth factor homologous factor

FT

flowthrough

GO

gene ontology

HA

hemagglutinin

ID

intercalated disc

i.p.

intraperitoneal injection

KEGG

Kyoto Encyclopedia of Genes and Genomes

RMP

resting membrane potential

TMT

tandem mass tags

VGSC

voltage-gated sodium channel

VGSCs

voltage-gated sodium channels

WCL

whole cell lysate

WGA

wheat germ agglutinin

WT

wildtype