Fhf2 gene deletion causes temperature-sensitive cardiac conduction failure

Fever is a highly conserved systemic response to infection dating back over 600 million years. Although conferring a survival benefit, fever can negatively impact the function of excitable tissues, such as the heart, producing cardiac arrhythmias. Here we show that mice lacking fibroblast growth factor homologous factor 2 (FHF2) have normal cardiac rhythm at baseline, but increasing core body temperature by as little as 3 °C causes coved-type ST elevations and progressive conduction failure that is fully reversible upon return to normothermia. FHF2-deficient cardiomyocytes generate action potentials upon current injection at 25 °C but are unexcitable at 40 °C. The absence of FHF2 accelerates the rate of closed-state and open-state sodium channel inactivation, which synergizes with temperature-dependent enhancement of inactivation rate to severely suppress cardiac sodium currents at elevated temperatures. Our experimental and computational results identify an essential role for FHF2 in dictating myocardial excitability and conduction that safeguards against temperature-sensitive conduction failure.

F ever-induced arrhythmias 1 and seizures 2 are well documented, and are often associated with mutations in sodium channels, suggesting that deficient sodium current reserve is an important determinant for electrical instability during hyperthermic states. Indeed, elevation in core body temperature by fever or external heating 3 is a known trigger for ventricular fibrillation/malignant syncope in patients with Brugada syndrome (BrS) 4 , an inherited arrhythmia condition diagnosed by characteristic electrocardiographic (ECG) abnormalities in the right precordial leads. Loss of function mutations in SCN5A, which encodes the pore-forming subunit of the cardiac sodium channel Na V 1.5, have been identified in B20% of BrS patients 5 . Biophysical analyses of mutant sodium channels from febrile BrS patients have not identified a unifying mechanism for the temperature-sensitive phenotype 6,7 . These data suggest that factors in addition to SCN5A are playing important roles in regulating the sodium current that ultimately predispose BrS patients to fever-induced arrhythmias.
FHFs, also termed iFGFs, are a family of proteins that bind to the cytoplasmic tails of voltage-gated sodium channels (VGSCs) [8][9][10] , modulating channel inactivation and cellular excitability [11][12][13] . We generated mice lacking fibroblast growth factor homologous factor 2 (Fhf2 KO ) to study its role in regulating cardiac excitability under normal and pathological states. Fhf2 KO mice have normal cardiac rhythm at baseline, but exhibit temperature-sensitive electrocardiographic changes, including coved-type ST elevations and progressive conduction failure that is fully reversible upon return to normal body temperature. Optical mapping reveals severe conduction slowing in mutant hearts at 37°C that is further exacerbated by temperature elevation. FHF2-deficient cardiomyocytes generate action potentials upon current injection at 25°C but are unexcitable at 40°C. Loss of FHF2 results in a hyperpolarizing shift of steady-state inactivation of the sodium current and accelerates the rate of closed-state and open-state sodium channel inactivation, which synergizes with temperature-dependent enhancement of inactivation rate to severely suppress cardiac sodium currents at elevated temperatures. Our experimental and computational results demonstrate that FHF2 is a key regulator of myocardial excitability, protecting the heart against conduction failure under hyperthermic conditions.

Results
Derivation and validation of Fhf2 KO mice. We engineered mice bearing a deletion within the X-linked Fhf2 gene (Fig. 1a,b) in order to test for possible effects on cardiac rhythm. The absence of FHF2 protein in Fhf2 KO mice was confirmed in immunoblots of heart and brain tissue (Fig. 1c), and by ventricular myocyte immunofluorescence (Fig. 1d). Notably, there was no noticeable change in Na V 1.5 protein levels or subcellular localization in Fhf2 KO ventricular myocytes. The IRES-lacZ insertion within the Fhf2 KO allele (Fig. 1a) allowed for whole-mount X-Gal staining of the heart and demonstrated widespread Fhf2 gene expression in atria, ventricles, and the His-Purkinje system (Fig. 1e).
Fhf2 KO  and functional assessments by transthoracic echocardiography were also normal under euthermic conditions (Supplementary  Table 2). However, Fhf2 KO/Y mice were highly temperaturesensitive. Elevation of core body temperature by external heat source resulted in marked conduction slowing as evidenced by progressive P and QRS wave prolongation and atrioventricular (AV) block ( Fig. 2a and Table 3). In contrast, Fhf2 WT/Y mice were able to tolerate high dose flecainide (30 mg kg À 1 ), and despite having marked prolongation of P, PR and QRS durations, the conduction parameters remained stable during temperature ramp up to 43°C ( Fig. 2d and Supplementary Fig. 1).
Isolated Fhf2 KO hearts retain temperature-sensitive defects. To rule out extra-cardiac contributions to the temperature-sensitive phenotype, including autonomic effects, explanted Fhf2 KO/Y and Fhf2 WT/Y hearts were Langendorff-perfused and subjected to a temperature ramp protocol with a maximum perfusion temperature of 43°C (Fig. 2e). Similar to the in vivo findings, Fhf2 KO/Y hearts exhibited progressive P and QRS wave prolongation, and AV block on volume-conducted ECG with temperature elevation. Coved-type ST elevations with T wave inversions were again seen in mutant hearts at temperatures greater than 40°C. Fhf2 WT/Y hearts displayed stable conduction parameters throughout the temperature recordings. Optical mapping of Fhf2 KO/Y and Fhf2 WT/Y hearts using a voltage-sensitive dye was performed to evaluate differences in epicardial conduction velocities (CV) at 37 and 39°C. Activation maps of the anterior wall were obtained at a pacing cycle length of 200 ms (Fig. 2f). Epicardial CV were markedly slower in Fhf2 KO/Y compared with Fhf2 WT/Y hearts at 37°C (0.25 ± 0.0063 m s À 1 versus 0.56±0.01 m s À 1 , KO versus WT, respectively P ¼ 1.35E-5, Student's t-test). CV slowing was global and not confined to the right ventricular outflow tract, as has been reported in some patients with BrS (refs [14][15][16]. Fhf2 KO/Y hearts showed further reduction in epicardial CV at 39°C, while Fhf2 WT/Y hearts maintained stable epicardial CV measurements at higher temperature. Fhf2 KO cardiomyocytes have reduced excitability at 40°C. To study the effects of FHF2 on cellular excitability, we acutely dissociated ventricular myocytes prepared from Fhf2 WT/Y and Fhf2 KO/Y mice and tested their ability to generate action potentials (AP) at 25 and 40°C in response to current injection through whole-cell patch. Fhf2 WT/Y cells could generate action potentials at either temperature, although AP waveforms were significantly attenuated at 40°C (Fig. 3a and Supplementary  Table 4). Loss of Fhf2 altered sodium channel inactivation in several ways. Channels in Fhf2 KO/Y cells underwent steady-state inactivation at more hyperpolarized potential than did channels in Fhf2 WT/Y cells at 25°C ( Fig. 3c and Supplementary Table 4). Fhf2 deficiency also accelerated the rate of sodium current decay when recorded at either 25 or 30°C, with mutation and elevated temperature combining to produce the fastest rate of open-state inactivation ( Fig. 3d and Supplementary Table 4). Sodium channel closed-state inactivation rates were assayed by comparing peak sodium conductance generated by voltage ramps to conductance generated by instantaneous voltage step. While longer ramp durations reduced peak sodium conductance under all conditions due to closed-stated channel inactivation, higher temperature and Fhf2 mutation each enhanced the effect of ramp duration on channel availability, with Fhf2 mutation and elevated temperature combining to yield the most severe reduction in availability (Fig. 3e, Supplementary Fig. 2 Table 4).

and Supplementary
Sodium channel inactivation kinetics were further studied in human embryonic kidney (HEK) cells transfected to express cardiac sodium channel Na v 1.5 with or without FHF2VY, the principal protein isoform of FHF2 expressed in mouse heart 13 . Patched HEK cells more easily survived temperature ramping than did cardiomyocytes allowing assessment of Na v 1.5 inactivation parameters at 25, 35 and 40°C often within the same cell. Steadystate voltage dependent Na v 1.5 inactivation at 25°C occurred with V 1/2 ¼ À 82.2 ± 1.6 mV (n ¼ 7) in the presence of FHF2VY and V 1/2 ¼ À 93.5±0.9 mV (n ¼ 9) in the absence of FHF2 (Po0.0002, Student's t-test) (Supplementary Table 5). The absence of FHF2VY also increased the rate of Na V 1.  Table 5). Most notably, even a 45 mV ms À 1 voltage ramp at 40°C caused a greater than 50% reduction in Na v 1.5 peak conductance in the absence of FHF2, while more than 80% of channels remained available under these conditions in the presence of FHF2VY ( Fig. 3f and Supplementary Fig. 3).
Computational models of Fhf2 WT and Fhf2 KO cardiomyocytes. To establish whether hyperthermic excitation failure is attributable to altered cardiac sodium channel inactivation gating, we modified a previously generated mouse ventricular cardiomyocyte computational model 17 to reflect the differential Na v gating properties in the presence and absence of FHF2. The Fhf2 WT and Fhf2 KO myocyte models differed only in the openand closed-state inactivation rate constants of the embedded Na v 12-state Markov model. Both models generated the same peak Na v conductance in response to step depolarization and the same voltage dependence of Na v activation ( Supplementary Fig. 4). But in comparison to the Fhf2 WT model, Na v conductance in the Fhf2 KO model displayed a hyperpolarizing shift in the voltage dependence of steady state inactivation ( Supplementary  Fig. 4), exhibited faster closed-state inactivation in voltage ramp simulations (Fig. 4a), and faster open-state inactivation (Fig. 4b). Analogous to recorded channels, inactivation rates were temperature-dependent, such that Fhf2 mutation and temperature elevation combined generated the fastest Na v inactivation rates (Fig. 4a,b, Supplementary Fig. 4). Current injection simulations (Fig. 4c) demonstrated the consequences of altered Na V inactivation gating. The Fhf2 WT model fired action potentials at both 25 and 40°C, although AP waveforms were attenuated at higher temperature, while the Fhf2 KO model exhibited complete excitation failure at elevated temperature.

Discussion
Preservation of cardiac sodium current density is critical for survival. Here we demonstrate that FHF2 acts to dampen the temperaturedependent acceleration of Na V  Although the ECG parameters of Fhf2 KO mice were normal at 37°C, the degree of conduction slowing measured by optical mapping was substantial. This difference may reflect that in vivo ECGs, unlike the optical maps, were measured during sinus rhythm, where near simultaneous multi-site activation of the ventricular myocardium by the His-Purkinje system can mask myocardial conduction slowing. Purkinje myocytes express additional FHF isoforms 18 , which may confer some degree of protection against conduction slowing within the specialized conduction system in Fhf2 KO mice. The full extent of conduction slowing observed in the isolated heart preparations may also reflect the increased dependence of sodium current density on the slope of initial membrane depolarization in Fhf2 KO myocytes. In the intact heart, action potentials are triggered following non-instantaneous depolarization mediated by gap junctional currents passing from upstream cells 19 . Therefore, sodium current density in Fhf2 KO hearts should be exquisitely sensitive to conditions that increase intercellular resistivity between cardiomyocytes, such as the increase in interstitial volume that may be seen with Langendorff perfusion 20 . Indeed, the observation that FHF2 may modulate the dependency of sodium current density on junctional conductance has important implications for arrhythmia mechanisms associated with conditions that produce intercellular uncoupling, such as acute ischaemia 21 or pathologic gap junction remodelling 22 .
The dynamic nature of the ECG abnormalities in FHF2 KO hearts is highly reminiscent of Brugada syndrome, where ECG parameters change from normal to coved-type ST elevations in the presence of fever or sodium channel blocking drugs. While the electrophysiological basis of the Brugada pattern ECG is controversial, some have reported focal conduction slowing in the right ventricular outflow tract as a potential mechanism [14][15][16] . Although optical mapping of Fhf2 KO hearts demonstrated global conduction slowing, additional analysis may be revealing. Furthermore, exploring the interplay between FHF2-dependent alterations in sodium current physiology and other ionic currents that have been implicated in the Brugada pattern ECG will certainly be of significant interest.
It should also be noted that in contrast to our results, Puranam et al. 23 recently reported embryonic lethality in their lineage of Fhf2 KO mice. While the basis for this discrepancy is unclear, one possibility is that this other reported mutant allele was maintained on a C57Bl/6 background, with background strain influencing phenotype. Whether this lineage of Fhf2 KO mice survives to adulthood in other background strains should be evaluated.
In summary, our results identify the critical role of FHF2 in maintaining adequate sodium current reserve in response to hyperthermic stress. These results have direct implications for fever-induced Brugada syndrome, where loss of function sodium channel mutations may conspire with FHF-dependent mechanisms, such as allelic expression levels, to produce temperature-sensitive effects. It will be of interest to extend our studies in Fhf2 mutant mice to additional fever models, such as challenge with lipopolysaccharide 24 or cytokines 25,26 . It is also worth noting that the co-expression of FHFs and Na V s in the heart and brain may point to a unifying theory for both febrile arrhythmias and seizures 23 .

Methods
Fhf2 KO mouse derivation. All protocols conformed to the Association for the Assessment and Accreditation of Laboratory Animal Care and the NYU School of Medicine Animal Care and Use Committee. A murine embryonic stem cell line bearing a 'knockout first with conditional potential' cassette integrated into the Fhf2 locus located on chromosome Xq26 (International Mouse Phenotype Consortium, clone EPD0339_4_F09) was injected into blastocysts to derive chimeric mice, which were outbred to establish viable mice bearing the targeted allele. Pronuclear injection of Cre recombinase-expressing plasmid into fertilized eggs of this lineage yielded progeny bearing Cre-mediated excision of Fhf2 coding exon 3, which specifies an integral portion of the FHF b-trefoil fold 10 . The Cre-excised allele specifies bicistronic transcripts encoding truncated FHF2 peptides (from exons 1 to 2) and E. coli LacZ (Fig. 1a). A 3-primer polymerase chain reaction reaction with Phyre DNA polymerase allows simultaneous detection of wild-type (WT) and knockout (KO) alleles (Fig. 1a,b). Mice carrying the KO allele were backcrossed three generations to the 129/svPas strain before use in all cardiac physiology experiments. Immunoblotting. Brain and exsanguinated heart tissues were homogenized in cold detergent-free buffer, after which membranes were solubilized by adding 1% Triton X-100. Clarified lysates were run on 4-20% precast polyacrylamide gradient gels Figure 2 | Hyperthermia-induced electrocardiographic (ECG) changes and conduction block in the Fhf2 KO/Y heart. (a) Representative surface ECG traces of adult (8-12 wk) Fhf2 WT/Y and Fhf2 KO/Y mice subjected to hyperthermia-induction protocol. External heat lamp was used to raise core body temperature from 37 to 43°C followed by cooling to 37°C (1°C min À 1 ). Fhf2 KO/Y ECG parameters were similar to Fhf2 WT/Y animals at 37°C (n ¼ 10) but were significantly prolonged in mutant mice at elevated temperatures (Z40°C) (n ¼ 9 mutants and 7 controls). Coved-type ST elevations and high degree AV block with periods of ventricular asystole were noted in Fhf2 KO/Y mice above 40°C. Fhf2 WT/Y mice did not exhibit altered ECG parameters at any temperature (37-43°C). QRS duration is plotted against core body temperature. (b) Representative intracardiac electrogram traces during hyperthermia-induction protocol. Intracardiac conduction intervals (AH, HV, AVI) of Fhf2 KO/Y were similar to Fhf2 WT/Y animals at 37°C but were significantly prolonged in mutant mice (n ¼ 3) at elevated temperatures (Z40°C). HV interval duration is plotted against core body temperature. Voltage ramp rate (log 10 mV ms -1 )  (b) Cardiomyocyte Na V peak conductance (gNa V -peak). gNa V -peak following À 120 mV to À 35 mV depolarization step was measured for Fhf2 WT/Y and Fhf2 KO/Y cells at 25°C and 30°C. (c) Cardiomyocyte Na V steady-state inactivation. Available gNa V at 25°C after 60 ms conditioning at À 140 to À 40 mV is expressed as fraction maximal gNa V . Vertical dashed lines indicate V 1/2 inactivation in Fhf2 WT/Y and Fhf2 KO/Y cells. (d) Cardiomyocyte Na V open-state inactivation rate. Superimposed representative traces (left) from Fhf2 WT/Y (black line) and Fhf2 KO/Y (grey line) cardiomyocytes at 25°C (top) and 30°C (bottom) following depolarization to À 35 mV. Fast exponential time constant (t) for decay of sodium current in response to voltage step to À 35 mV are plotted for Fhf2 WT/Y and Fhf2 KO/Y cells at 25 and 30°C (right). (e) Voltage ramp Na V inactivation in cardiomyocytes. gNa V -peak in response to voltage ramps to À 50 mV at different rates was expressed as percentage of gNa V -peak elicited by step depolarization (N ramp rate). Slower ramp rates decrease gNa V -peak through closed-state inactivation, which is sensitized by both elevated temperature and Fhf2 mutation (histogram). See Supplementary Fig. 2 for representative conductance traces. (f) Voltage ramp Na V 1.5 inactivation with and without FHF2VY. HEK cells transfected with Na V 1.5 ± FHF2VY subjected to voltage ramps to À 30 mV at different rates at 25, 35 and 40°C. Na V 1.5 closed-state inactivation is increased by slowed voltage ramp, temperature elevation and the absence of FHF2VY. See Supplementary Fig. 3  (Invitrogen) and transferred to nitrocellulose (Bio-Rad) overnight at 4°C. Nitrocellulose membranes were incubated in blocking buffer consisting of PBS (Na V 1.5) or TBS (FHF2) with Tween-20 (0.05%) and 5% nonfat dry milk. Membranes were then incubated with specific primary antibodies diluted in 5% nonfat dry milk in PBST/TBST (0.05%) overnight at 4°C followed by wash steps and secondary antibodies (Li-Cor). Antigen complexes were visualized and quantified with the Odyssey Imaging System (Li-Cor). The uncropped western blot images shown in Fig. 1c are displayed in Supplementary  Fig. 5.
Immunocytochemistry. Dissociated ventricular myocytes were fixed in 4%PFA for 5 min. Cells in suspension were spun onto slides using Shandon Cytospin II (Speed 250 r.p.m., 2 min). Cells were blocked with 10% serum and 0.01% Triton in PBS for 1 h, and then incubated with primary antibodies overnight. Cells were then washed in PBS and incubated with secondary antibodies conjugated to Alexa Fluor dyes (Invitrogen) for 1 h before mounting. Slides were coverslipped with Vectashield mounting media with DAPI (Vector Laboratories). Stained cells were visualized using a Leica TCS SP5 confocal microscope with Leica LAS AF acquisition software.
Mouse in vivo electrophysiology and flecainide drug challenge. ECGs were obtained using subcutaneous electrodes attached at the four limbs (MP100, BIOPAC Systems). 8-12 week-old, male mice were anaesthetised with inhaled (2% v/v) isoflurane. Heart rate and core body temperature (rectal temperature probe) were continuously recorded. Flecainide (15 mg kg À 1 or 30 mg kg À 1 ) was Voltage ramp rate (log 10 mV ms -1 ) 100 ms current injections ranging from 0 to 600 pA (insets) were applied to Fhf2 WT and Fhf2 KO models at 25 and 40°C. Both models are excitable at 25°C, but only the Fhf2 WT model is excitable at 40°C, comparable to cardiomyocyte recordings (Fig. 3a).
administered via intraperitoneal injection. ECG analysis was performed in an unbiased fashion where 100 beats at each temperature endpoint were analysed using LabChart 7 Pro version 7.3.1 (ADInstruments, Inc). Detection and analysis of P wave, PR interval and QRS wave intervals were set to Mouse ECG parameters. Intracardiac recordings were obtained in 8-12 week-old, male mice under isoflurane anaesthesia (1.5% v/v) using an octapolar catheter (EPR-800, Millar Instruments) placed via the right internal jugular vein for bipolar recordings at the level of the His bundle. Intracardiac intervals were measured from the His bundle catheter when the atrial, His, and ventricular electrograms were stable over 425 beats. Measurement of the AH interval was taken from the His bundle recording from the onset of deflection from baseline of the local atrial electrogram to the onset of deflection from baseline of the His bundle electrogram. Measurement of the HV interval was taken from the onset of deflection from baseline of the His bundle electrogram to the onset of deflection from baseline of the earliest ventricular electrogram, whether on surface ECG or intracardiac recordings. Core body temperature was initially maintained at 37°C and gradually elevated up to 43°C at a rate of 1°C min À 1 using an external heat lamp.
Transthoracic echocardiography. Echocardiography was performed using the Vevo 2,100 high-resolution ultrasound imaging system with a real-time 30 MHz linear array scanhead (MS400) at a frame rate of 235 fps, a focal length of 8 mm, and a 10 Â 10-mm field of view (Visualsonics; Toronto, Canada). 8-12 week-old, male mice were anaesthetised with 2% isofluorane, and hair was removed from the chest using a depilatory cream (Nair; Church & Dwight Co, Inc; Princeton, NJ). Warmed ultrasound transmission gel was placed on the chest and used to obtain left ventricular endpoints of cardiac function. B-mode cardiac imaging was conducted on transverse (short axis) plane. The papillary muscles were used for the short axis imaging landmark. M-mode recordings of the left ventricle were also recorded at the short axis B-mode imaging plane to obtain left ventricular function and dimensions through the cardiac cycle. Heart rate was monitored and core body temperature was maintained at 37.5°C using a heated platform and a hair dryer throughout the procedure. Data analysis was performed on VisualSonics Vevo 2,100 V1.5.0 software (Visualsonics; Toronto, Canada). The following parameters were measured using three cardiac cycles from short axis M-mode images: diastolic and systolic left ventricular internal diameter, anterior wall thickness, and posterior wall thickness. From these measurements, left ventricular ejection fraction, per cent fractional shortening, stroke volume, and cardiac output were calculated within the Vevo software.
Optical mapping. High-resolution optical mapping experiments were performed as follows: excised hearts from 8 to 12 week-old, male mice were initially perfused with Tyrode's solution to clear blood and stabilize the heart, followed by Tyrode's solution containing 10 mM blebbistatin. Hearts were allowed to recover for 20 min and then stained with the voltage-sensitive dye, Di-4-ANEPPS (Molecular Probes Inc., Eugene, OR, USA). Light from green LEDs (530 nm; ThorLabs) was used as an excitation source and the emitted light (620 nm long pass) was detected with 1 high-resolution CMOS camera (Mi-CAM Ultima-L;SciMedia) at 1,000 frames per s in bin mode (100 Â 100 pixels) with 14-bit resolution. Images were processed using a custom software package.
Action potential recordings from adult cardiomyocytes. Excitability of acutely dissociated ventricular cardiomyocytes from FHF2 WT/Y and FHF2 KO/Y mice was conducted using a MultiClamp700 Amplifier, Digidata 1,440 analogue/digital converter and Clampex10 software (Molecular Devices). Cells were placed in the recording chamber under a Nikon Eclipse microscope and perfused with carbogen-bubbled bath solution (115 mM NaCl, 26 mM NaHCO 3, 3 mM KCl, 1.2 mM KH 2 PO 4 , 3 mM glucose, 2 mM myoinositol, 2 mM Na pyruvate, 7 mM HEPES, 1.2 mM Mg 2 SO 4 , 2 mM CaCl 2 , 0.2 mM CdCl 2 at pH 7.2) maintained at 25°C using an in-line heater (Warner Instruments). Patch pipettes were pulled with a P97 Micropipette Puller (Sutter Instruments) to yield a resistance of 1-1.5 MO when filled with 120 mM K gluconate, 4 mM NaCl, 5 mM KOH-buffered HEPES, 5 mM KOH-buffered EDTA, 15 mM glucose, 1 mM MgSO 4 , 3 mM Mg-ATP, 0.1 mM Na-GTP at pH 7.2. Pipette approach to cell was visualized using a 60 Â water immersion lens and infrared illumination with differential interference contrast. Following tight seal formation and break-in to achieve whole-cell access, voltage clamp depolarization steps were used to confirm presence of large inward sodium current and measure capacitive current transients before switching to current clamp mode. Steadily applied negative current was used to set the amplifier-measured voltage to À 85 mV, which was equal to À 95 mV membrane potential due to a 10 mV junctional potential between pipette and bath solutions. Excitability was assessed by applying 200 ms injection sweeps of continuous current ranging from 0 to 480 pA in 40 pA steps. Membrane voltage recordings were high-pass filtered at 10 kHz and digitally acquired at 50 kHz. After data acquisition, the bath temperature was ramped to 40°C over a 3-5 min time span, and excitability was retested with the same protocol. Only cells for which input resistance deviated less than 20% during temperature ramp were included in analysis.
Na v 1.5 currents in HEK cells. 293T human embryonic kidney cells previously described in 28 , and displaying little or no endogenous sodium channel expression 29 were transiently cotransfected with a 2:1 mixture of Na v 1.5-expressing plasmid 30 , and a pIRES2-ZsGreen bicistronic plasmid (Clontech) expressing ZsGreen and mouse FHF2VY proteins. The same pIRES2-ZsGreen plasmid without FHF2 coding sequence served as control. Cells were trypsinized 3 h post-transfection, seeded onto gelatinized coverslips, and were used for recording after 48 h. For sodium current recordings, coverslips were transferred to recording chamber containing carbogen-buffered bath solution (115 mM NaCl, 26 mM NaHCO 3 , 3 mM KCl, 10 mM glucose, 4 mM MgCl 2 , 2 mM CaCl 2 , 0.2 mM CdCl 2 , 3 mM myoinositol, 2 mM Na pyruvate, 7 mM NaOH-buffered HEPES pH 7.2) at 25°C and green fluorescent cells were whole-cell patched with pipettes filled with 104 mM CsF, 50 mM tetraethylamine chloride, 10 mM HEPES pH 7.2, 5 mM glucose, 2 mM MgCl 2 , 10 mM EGTA, 2 mM ATP, 0.2 mM GTP and having 1-2 MO resistance. For all recording protocols, sodium current was isolated during data acquisition by P/N subtraction of leak and capacitive currents (N ¼ À 6). To ensure voltage clamping during sodium channel activation was adequate, cells were subjected to a 19-sweep series of voltage steps from a hold of À 120 mV to between À 80 mV and þ 10 mV in 5 mV increments. As criterion for adequate clamp, transient current peaks for all voltage commands were nested within the larger current trace of a following or preceding voltage step command.
Steady-state channel inactivation protocol. To determine the voltage dependence of steady state channel inactivation, a 19-sweep protocol used À 120 mV holding command, a 60 ms variable test voltage step ( À 120 þ 5(n À 1) mV), followed by a À 25 mV reporting pulse.
Voltage ramp protocol. As a measure of closed state channel inactivation rate, a 10-sweep protocol used À 120 mV hold command followed by depolarization to À 30 mV either instantaneously (voltage step) or as a ramp ranging in time from 2 ms ( ¼ 45 mV ms À 1 ) to 18 ms ( ¼ 5 mV ms À 1 ).
Temperature elevation. After recording at 25°C, temperature was ramped to 35°C and later to 40°C. At elevated temperatures, the cell was first tested as above to ensure maintenance of tight clamp, after which the voltage ramp protocol was conducted. For many cells, voltage ramp protocols could be successfully run at all three temperatures (examples in Supplementary Fig. 3).
Sodium current recordings in adult cardiomyocytes. All I Na recordings in isolated cardiomyocytes were conducted in whole-cell configuration at 25 or 30°C. Recording pipettes were filled with a solution containing (in mM): NaCl 5, CsF 135, EGTA 10, MgATP 5, HEPES 15, pH 7.2 with CsOH. Cells were maintained in a solution containing (in mM): NaCl 5, CsCl 112.5, TEACl 20, CdCl 2 0.1, MgCl 2 1, CaCl 2 , 1, HEPES 20, Glucose 11, pH 7.4 with CsOH. To determine the peak current voltage relation, 200 ms voltage pulses were applied to V m À 90 mV to þ 30 mV in 5 mV voltage steps, from a holding potential of V m ¼ À 120 mV. Interval between voltage steps was 3 s. Steady state inactivation was determined by stepping V m to conditioning voltages of between À 130 mV and À 20 mV for 60 ms, followed by a 30 ms test pulse to V m ¼ À 20 mV to elicit I Na . The steady state voltage-dependent inactivation curves were fitted to Boltzmann's functions. As a measure of closed state channel inactivation rate, a 10-sweep protocol used À 140 mV hold command followed by depolarization to À 50 mV either instantaneously (voltage step) or as a ramp ranging in time from 2 ms ( ¼ 45 mV ms À 1 ) to 18 ms ( ¼ 5 mV ms À 1 ). All recordings were obtained utilizing an Axon multiclamp 700B Amplifier coupled to a pClamp system (versions 10.2, Axon Instruments, Foster City, CA).
Cardiomyocyte excitation parameters. Spike threshold was defined as the point of accelerated upward voltage deflection at the base of a spike, mathematically defined as the voltage when d 2 V/dt 2 40 as determined using Clampfit software (Molecular Devices). where Dq is the current transient time integral to a negative voltage step command DV. Leak conductance was measured as where DV m is the steady-state change in membrane voltage in response to a negative current clamp command I step .
where E Na is the sodium reversal potential and I NavPeak is the peak inward current in response to a voltage command V com within the linear ohmic range of the current/voltage relationship.
Voltage dependence of steady state sodium channel inactivation. Peak sodium current accompanying the À 25 mV reporting pulse was measured at each test voltage, and the fraction of channels available (not inactivated) at test voltage equalled I Na-peak (V test )/I Na-peak (maximum). To obtain V 1/2 and k values for voltage dependence of steady state inactivation, current data were fitted to Boltzmann equation: Open-state sodium channel inactivation kinetics. Transient sodium current decay from 90% current peak to baseline was fitted to a two-term exponential decay function where I 1 and I 2 are the fast and slow decaying components of the transient current.
The fast component I 1 associated with decay constant t fast constituted between 70 and 100% of the total current in all analysed cells.
Closed-state sodium channel inactivation kinetics. Data from voltage ramp protocols provided an estimate of closed-state channel inactivation rate as a function of voltage ramp rate. For each ramp command at each time point, we calculated g Nav ¼I Nav = V com À E Na ð Þ ð 7Þ yielding sodium conductance traces as shown in Supplementary Fig. 3. Similarly, g NavPeak ¼I NavPeak = V comðatPeakÞ À E Na À Á ð8Þ from which was calculated % maximal g NavPeak ramp ð Þ¼100Âg NavPeak ramp ð Þ=g NavPeak step ð Þ ð9Þ Computational modelling. The Fhf2 WT and Fhf2 KO cardiomyocyte models were generated by introducing modifications to the murine ventricular cardiomyocyte model from Bondarenko et al. 17 . These changes included (1) adjusting membrane capacitance to 70 pF to more closely match our recorded measurements, (2) recalibrating the 'leak' conductances by dividing the inward rectifying postassium conductance two-fold and replacing with an equal density of potassium conductance to achieve nonspiking responses to current injections similar to cardiomyocyte recordings (Fig. 4c), (3) changing the densities of ultrarapid-activating delayed rectifier and rapid-activating delayed rectifier potassium conductances to 0.112 mS mF À 1 and 0.234 mS mF À 1 , respectively, in order to obtain more realistic post-spike repolarization and (4) replacing the fast sodium conductance with a 12-state Markov model sodium conductance: Parameter values shared by both Fhf2 WT and Fhf2 KO models are Aa ¼ 2.44375 ms À 1 , Ab ¼ 0.01325 ms À 1 , Ag ¼ 150 ms À 1 , Ad ¼ 40 ms À 1 , Va ¼ Vb ¼ 6 mV, n 1 ¼ 5.422, n 2 ¼ 3.279, n 3 ¼ 1.83, n 4 ¼ 0.738, a ¼ (O on /C on )^0.25, b ¼ (O off /C off )^0. 25. For the Fhf2 WT model, AC on ¼ 0.001 ms À 1 , AC off ¼ 0.108 ms À 1 , AO on ¼ 1.05 ms À 1 , AO off ¼ 0.003 ms À 1 , and V shift ¼ À 58 mV, while in the Fhf2 KO model AC on ¼ 0.054 ms À 1 , AC off ¼ 0.027 ms À 1 , AO on ¼ 1.5 ms À 1 , AO off ¼ 0.001 ms À 1 , and V shift ¼ À 61 mV. These parameter differences preserved the voltage dependence of Na V activation in the models while causing a 7 mV hyperpolarizing shift in voltage dependence of inactivation for the Fhf2 KO model ( Supplementary Fig. 4) and accelerating closed-state and open-state inactivation rates (Fig. 4a,b) compared with the Fhf2 WT model. Voltage-clamp and current-clamp simulations were run using the Myokit software interface 32 and were initiated with a 200 ms channel state equilibration phase. Voltage-clamp simulations included single-step and double-step protocols to measure voltage dependence of Na V activation and inactivation along with a ramp protocol comparable to that employed for experimental recordings to assess closed-state inactivation. For current-clamp simulations, an initial constant negative current was applied to adjust membrane voltage to À 95 mV before adding positive current steps in multiples of 50 pA (0.7 pA pF À 1 ).
Statistics. Quantitative values are expressed as mean and standard error of the mean for each group. Two-tailed Student's t test was used to compare differences in continuous variables between mutant and control animals and cells. Po0.05 was considered statistically significant. Sample size calculations were done using preliminary data to design the experiment for measuring continuous variables. Groups were constructed to detect a 30% difference between experimental and control groups with a power of 90% and a significance level of 0.05. Experimental groups were blinded until the endpoints were analysed. Animal studies were done before genotyping ensuring blinded observations. The Prism statistical package (version 6, GraphPad) was used for analyses.