The efficacy of Ranolazine on E1784K is altered by temperature and calcium

E1784K is the most common mixed syndrome SCN5a mutation underpinning both Brugada syndrome type 1 (BrS1) and Long-QT syndrome type 3 (LQT3). The charge reversal mutant enhances the late sodium current (INa) passed by the cardiac voltage-gated sodium channel (NaV1.5), delaying cardiac repolarization. Exercise-induced triggers, like elevated temperature and cytosolic calcium, exacerbate E1784K late INa. In this study, we tested the effects of Ranolazine, the late INa blocker, on voltage-dependent and kinetic properties of E1784K at elevated temperature and cytosolic calcium. We used whole-cell patch clamp to measure INa from wild type and E1784K channels expressed in HEK293 cells. At elevated temperature, Ranolazine attenuated gain-of-function in E1784K by decreasing late INa, hyperpolarizing steady-state fast inactivation, and increasing use-dependent inactivation. Both elevated temperature and cytosolic calcium hampered the capacity of Ranolazine to suppress E1784K late INa. In-silico action potential (AP) simulations were done using a modified O’Hara Rudy (ORd) cardiac model. Simulations showed that Ranolazine failed to shorten AP duration, an effect augmented at febrile temperatures. The drug-channel interaction is clearly affected by external triggers, as reported previously with ischemia. Determining drug efficacy under various physiological states in SCN5a cohorts is crucial for accurate management of arrhythmias.


Ranolazine does not affect conductance. Raw current traces in
show the effects of 0 µM and 100 µM Ranolazine on WT and E178K at 0 nM and 2500 nM cytosolic calcium (only 34 °C shown). E1784K reduced (p < 0.0001) the peak current and conductance density compared to WT. Elevated temperature (34 °C) increased (p < 0.0001) peak current and conductance density in WT but not in E1784K. Ranolazine had no effect (p > 0.05) on peak current or conductance density (Table 1 and Fig. 3B). Figure 3A shows normalized conductance plotted against the test potential at 34 °C. E1784K (p < 0.0001) and elevated temperature (p = 0.0003) depolarized the conductance midpoint (GV-V 1/2 ). The conductance slope (GV-z) was reduced (p < 0.0001) in E1784K compared to WT and increased (p < 0.0001) when temperature was elevated in both channel variants. The interaction between channel variant and temperature had no effect on GV-V 1/2 and GV-z. Normalized conductance was unchanged in all conditions with Ranolazine (Table 2).
E1784K availability is decreased in Ranolazine. Normalized current is plotted against membrane potential in Fig. 4. E1784K hyperpolarized (p < 0.0001) the SSFI midpoint (SSFI-V 1/2 ) compared to WT. Elevated temperature depolarized (p < 0.0001) SSFI-V 1/2 in both WT and E1784K. At 34 °C and 0 nM cytosolic calcium, SSFI-V 1/2 in E1784K was hyperpolarized (p < 0.0001) in 100 µM Ranolazine compared to WT ( Fig. 4C and Table 2). This effect was not significant at 2500 nM cytosolic calcium. Analogous to the shifts on GV-z, SSFI-z was decreased in E1784K and increased with elevated temperature (p < 0.0001). The slope was reduced in all conditions when Ranolazine was increased from 10 µM to 100 µM (p < 0.05, Table 2). Fast inactivation onset kinetics are not altered with Ranolazine. Fast inactivation onset kinetics at depolarized potentials (>−50 mV) were measured from τ on of the mono-exponential fits. E1784K fast inactivation onset kinetics were accelerated regardless of temperature (Fig. 5A,B). Onset kinetics were accelerated (decreased τ on ) with elevated temperature in WT compared to E1784K (p < 0.01). WT and E1784K onset kinetics were decelerated (increased τ on , p < 0.05) in Ranolazine as a function of voltage and cytosolic calcium at 22 °C (values reported in Table 3). These drug effects on τ on were not significant at elevated temperature. Figure 5C shows the Q 10 values at 0 µM and 10 µM Ranolazine for all conditions. We observed high variability in the temperature coefficient at −50 mV compared to other voltages. At −50 mV, both Ranolazine and cytosolic calcium mutually affect thermosensitivity in WT: Q 10 decreased at 0 nM cytosolic calcium and increased at 2500 nM cytosolic calcium in 10 µM Ranolazine. At more depolarized voltages than −50 mV, subtle alterations occurred in Q 10 (Fig. 5C). E1784K Q 10 was not sensitive to Ranolazine. However, E1784K thermosensitivity was dampened in cytosolic calcium at −50 mV compared to other voltages and to WT.

Ranolazine does not suppress thermosensitive late INa in E1784K with elevated cytosolic calcium.
Representative normalized late I Na current traces are shown in Fig. 6A,B at 0 µM and 100 µM Ranolazine (only 34 °C shown). Late I Na percent and density are shown in Fig. 6C,D as bar graphs. Late I Na percent and density in E1784K increased (p < 0.01) by 11-fold and 7-fold, respectively, with elevated temperature at 0 nM cytosolic calcium ( Fig. 6D and Table 4). This increase in late I Na was almost fully attenuated in 10 µM Ranolazine. Late I Na percent decreased in elevated cytosolic calcium (p < 0.01) but there was no effect on late I Na density in E1784K. Late I Na percent and density in E1784K were not suppressed with Ranolazine at 2500 nM cytosolic calcium ( Fig. 6D and Table 4).
Ranolazine does not enhance UDI in E1784K with elevated cytosolic calcium. Sustained or repetitive depolarizations induce slow inactivation in Na V 1.5, which was indirectly measured by the use-dependent inactivation (UDI) protocols described in the methods. We were unable to measure fast inactivation recovery kinetics in E1784K which may be decelerated by Ranolazine and contribute to UDI 54 . Use-dependence was measured at 1 Hz and 3 Hz, mimicking resting heart rate (60 bpm) and tachycardia (180 bpm), respectively. Normalized current plotted against time for UDI measured at 1 Hz and 3 Hz are shown in Fig. 7 (only 34 °C shown).
UDI plateau (y 0 ) was greater (p = 0.0430) at elevated temperature at both 1 Hz and 3 Hz, but the shift was larger in E1784K at 1 Hz (Table 5). y 0 decreased to different levels in Ranolazine (reported in Table 5). At high UDI frequencies, E1784K y 0 decreased in Ranolazine at 34 °C compared to WT (Fig. 7 shows only 34 °C, Table 5). Our statistical results suggest that the drug effects on UDI (3 Hz) in E1784K are limited in elevated cytosolic calcium (Table 5).

Figure 2.
Ranolazine docked to Na V 1.5-Na V Pas. The side view of Na V 1.5-Na V Pas homology model is shown docked to Ranolazine. The enlarged inset shows the cartoon structure of the drug binding to the central domains of the channel. The aromatic F1760 residue is outlined. Below the inset is a 3D-structure of Ranolazine (Nitrogen is blue, Oxygen is red, Carbon is green, and Hydrogen is grey). Conserved residues in DIII-DIV linker and CTD between Na V 1.5 and Na V Pas are indicated by a red asterisk.
Electrical restitution curves (ERCs) at 90% repolarization were constructed from plotting APD 90 against the diastolic interval as shown in Fig. 9A,B in endocardial cells. The last two beats were included in the ERCs to exemplify bifurcation and alternans-induction at critical diastolic intervals. The APD 90 for WT follows a similar trend to previously published ERCs, typifying a relatively stable APD rate dependence [56][57][58] . E1784K has a higher APD 90 in all cardiac cells, especially the mid-myocardium, compared to WT (not shown in Fig. 9). At 37 °C, the mutant causes bifurcation in APD 90 , mainly in epicardial cells, indicative of alternans (not shown); however, endocardial cells also experience alternans at febrile temperature in addition to epicardial cells (Fig. 9B). Upon drug perfusion, bifurcations were observed at higher diastolic intervals in E1784K (Fig. 9). The drug-induced bifurcations in ERCs were augmented at febrile temperature in all cardiac cells.
A linear relationship is established between the last two AP beats at each frequency step (shown in Fig. 9C), with no alternans. Divergence from linearity is indicative of alternans occurrence. At both 37 °C and 41 °C, WT cells had no alternans, even upon 10 µM Ranolazine perfusion, showing linearity with a slope = 1 (Fig. 9C). In drug-free conditions, E1784K had a linear relationship at high BCLs (low frequencies), but deviated from linearity beginning at intermediate BCLs; distortion in linearity is observed at lower frequencies in epicardial cells (Fig. 9C). This relationship in E1784K is augmented with febrile temperature. The prolonged APD 90 in E1784K were shortened with Ranolazine at very low frequencies, and alternans were quickly induced even during bradycardia, an effect exacerbated by febrile temperature (Fig. 9C).

Discussion
Our goal was to determine whether Ranolazine reduced channel dysfunction in E1784K under the triggering conditions of elevated temperature and cytosolic calcium. Ranolazine did not attenuate gain-of-function in E1784K when temperature and cytosolic calcium were elevated. Ranolazine has minimal effects on conductance in Na V 1.5 48,49 . The drug follows the modulated receptor hypothesis, targeting the open/inactivated states at depolarized potentials, thereby suppressing late I Na 59 . Physiological events, such as acidaemia, enhance Ranolazine antiarrhythmic effect by augmenting late I Na , thus providing the drug with a larger open-state channel substrate to target 48,60,61 . In addition to physiological modulators, SCN5a mutations often alter voltage-dependence of the channel, which modify drug effects on Na V 1.5. To date, Ranolazine has been screened against only ∆KPQ 50,52 , Y1767C 49 , R1623Q 51 , and D1790G 62 . Our study is the first to show the combined external triggers and SCN5a mutation effects on Ranolazine.
Ranolazine efficacy was enhanced at elevated temperature. Similar to acidosis effects, elevated temperature increases the late open probability in E1784K 48,53 . Late I Na percent and density increased by 11-and 7-fold, respectively, with elevated temperature. We previously reported a 3.54-fold increase in late I Na percent when temperature was elevated from 22 °C to 34 °C 41 ; however, we used CHOK1 cells to study E1784K thermosensitivity. The temperature coefficient (Q 10 ) partly depends on lipid-channel interactions in the membrane, which differ between heterologous expression systems 63,64 . The HEK293 lipid bilayer is less viscous than CHOK1 cells as observed in our whole-cell recordings, which may justify the heightened Q 10 in the late I Na measurements. At elevated stimulation frequencies and temperature, therapeutic Ranolazine decreased channel availability by increasing channel use-dependence in E1784K.
Although Ranolazine efficacy appears to be increased by temperature, its efficacy appears to be dampened by the combination of both elevated temperature and cytosolic calcium. At 34 °C, E1784K late I Na percent was depressed with elevated cytosolic calcium, but there was no effect on late I Na density. These opposing changes in late I Na percent and density may be attributed to the increased peak I Na density with elevated cytosolic calcium at 34 °C. Although not significant, the shift contributes to late I Na percent calculation ( their pharmacological analysis suggested drug-binding at hyperpolarized voltages 48 . Ranolazine exhibits partial preference to the closed state in Na V 1.5, as suggested by the hyperpolarizing shift in SSFI observed with Ranolazine. The shift mainly decreases channel availability near resting potential in cardiomyocytes. Other classic antiarrhythmics, anticonvulsants, and local anesthetics have similar effects, yet follow the modulated receptor hypothesis 59,66 . Sokolov et al. also report that the drug-induced block in late I Na occurs by a slow-mode recovery in slow inactivation, which is exacerbated at low pH 48 . This effect was mildly observed in this study. However, it is difficult to correlate the drug effects in E1784K UDI to late I Na , as the former had minor but significant shifts. Ranolazine shares a very similar structure with the class 1b antiarrhythmic drug lidocaine, which, like other sodium blockers, preferentially binds to F1760 and, to a lesser extent, Y1767 in DIVS6 49,52 . Ranolazine has high lipophilicity and can only bind to its receptor sites by traversing the phospholipid membrane and entering the central cavity through the inner vestibule. However, lateral pores, known as fenestrations, are alternative routes for large compounds like Ranolazine to access the central cavity. The fenestrations in the Na V Pas-Na V 1.5 model were unavailable for drug binding in auto-docking due to their constricted sizes 67 . It would be interesting to determine whether E1784K alters fenestration size in Na V 1.5, modifying drug entry via the fenestrations.
The interaction between Ranolazine and extracellular channel regions is unknown, but is unlikely due to its lipophilicity; a crystal structure of the channel/drug interaction would elucidate the drug-induced modifications in gating. The newly discovered aryl sulfonamide antagonists preferentially stabilize Na V 1.7 DIVS4 activation thereby stabilizing the fast inactivated state and suppressing late I Na 68 . Ranolazine may be structurally modified to include other moieties, like anionic aryl sulfonamides, for further optimizing its selectivity for targeting late versus peak I Na .
E1784K-induced Structural Rearrangements in NaV1.5 and its Impact on Ranolazine. We speculate E1784K affects fast inactivation via two possible mechanisms in Na V 1.5, thereby altering drug-channel interactions. Figure 1 shows the channel structures discussed, as follows: (1) E1784K hyperpolarizes the voltage-dependence of SSFI, thus stabilizing the interaction between the channel and the fast inactivation particle 7,11,12,22,28 . Fast inactivation onset is correlated with DIVS4 activation, whereas channel recovery is rate limited by charge immobilization of DIVS4 69,70 . The charge reversal mutant, E1784K, may enhance the transition of DIVS4 between closed and open states, as suggested by the Peters-Ruben model 43 . We postulate that this effect may be due to an electrostatic repulsion between the CTD mutant and conserved positive residues in DIVS4, given their close proximity 67 . This repulsion could make the DIVS4 in E1784K more mobile, which might explain the accelerated fast inactivation onset and recovery kinetics 22,37,41,43 . Fast inactivation kinetics in E1784K are not enhanced by temperature, so it does not seem justified to attribute the thermosensitive late I Na in E1784K to increased recovery kinetics. Rather a rearrangement may occur in DIVS4, conforming the voltage sensor to a state in which conductance in E1784K is higher 43 .  (2) E1784K alters the structure of CTD by disrupting the native hydrophobic and electrostatic interactions holding the EF-like hand domain (α 1 -α 4 ) tight with the IQ motif (α 6 ) 31 . Calcium sensitivity is imparted in Na V 1.5 via CaM, which binds to the IQ motif (α 6 ) via its C-lobe or N-lobe depending on cytosolic calcium levels [71][72][73] . During diastole or systole, CaM is calcified to different extents at its N-lobe 74 . Calcified CaM has a lower affinity for the IQ motif and binds, via its C-lobe, to DIII-DIV linker, forming a tripartite complex. This interaction is thought to prevent the DIII-DIV linker fast inactivation particle from occluding the pore, increasing channel availability near resting potential 71,73 . With depolarized potentials, the CaM C-lobe stabilizes the fast inactivation particle, suppressing late I Na , as in ∆KPQ and other mixed syndrome mutants 37,40,44 . Some studies refute the tripartite complex formation and favor a Ca V 1.2-like regulation of inactivation in Na V 1.5 74,75 . In those studies, the calcium-calmodulin complex is localized to CTD 76,77 . The Na V Pas structure showed intermolecular interactions between DIII-DIV linker, CTD, and DIVS4 67 .
Motoike et al. reported CaM-independent interactions between the inter-linkers in Na V 1.5 36 , suggesting that CaM acts as an auxiliary channel modifier during a calcium signal 75 . Calcium regulation in Na V 1.5 is mediated by CaM since the dual EF-like hand domains in CTD do not bind calcium 74,75 . In light of these structural models, we speculate that E1784K decouples both the calcium-dependent and calcium-independent interactions between the DIII-DIV linker and the CTD. Thus, E1784K inhibits calcium-dependent facilitation in Na V 1.5. We propose that the decoupling in CTD caused by E1784K may create a high entropy, unstable structure. Upon a calcium signal, the calcified calmodulin has reduced affinity for the IQ motif, thus augmenting CTD entropy 73 .  The pulse protocol is identical to that used to measure channel conductance (refer to Methods). Panel C includes Q 10 coefficient values for all conditions between −50 mV to +10 mV. Cytosolic calcium seems to modulate Ranolazine effects on WT Q 10 , as elevated cytosolic calcium heightened thermosensitivity. Cytosolic calcium made −50 mV fast inactivation onset in E1784K less thermosensitive, consistent with decoupling between CTD and Domain III-IV linker mechanism, explained in the discussion. Both mechanisms (1) and (2) may occur simultaneously in E1784K. The calcium effects in Na V 1.5 are localized to CTD. No reports have shown direct interaction between calcium-calmodulin and DIVS4, so if mechanism (2) occurs, it may be via an indirect effect on DIVS4.
In light of the discussed structural insights, we speculate that Ranolazine can easily access the inner vestibule with non-calcified calmodulin, since the molecule binds tightly to the IQ motif. Ranolazine efficacy, however, is hampered by cytosolic calcium, suggesting an interaction between the drug and the channel at CTD. The high entropy CTD in calcified calmodulin seems to physically hinder Ranolazine from entering into the inner vestibule.
Physiological and Medical Implications. Elevated temperature and cytosolic calcium are two of many other physiological triggers that occur during exercise and are common to other pathophysiological states, such as myocardial ischemia or infarction, and heart failure 78,79 . The majority patients with SCN5a mutations show ameliorated LQT3 phenotype during exercise 80 . Functional studies have correlated this to a stimulation frequency or calcium-induced reduction in late I Na 44 . However, it is clear from our study, focusing on E1784K, that the SCN5a mutant response to triggers can be unique 37,41 . Thus, it is necessary to study antiarrhythmics in SCN5a cohorts during different physiological states as the mutant-trigger effect may determine drug efficacy.
Our AP simulations clearly show pro-arrhythmic effects of Ranolazine, which are exacerbated by febrile temperatures. Electrical restitution curves clearly show a critical diastolic interval at which alternans are triggered. Our AP simulations provide evidence of Ranolazine's arrhythmogenicity, as it does not shorten APD 90 in E1784K at high heart rates. At low heart rates and at body core temperature, the drug shortens APD 90 in cardiac cells. However, with normal and elevated heart rates, the drug induces alternans, an effect exacerbated at febrile temperature. The critical diastolic intervals at which alternans are caused by the drug appear earlier (at higher BCLs) at febrile temperature.
E1784K induces alternans with higher prevalence in epicardial cells at low heart rates. This result coincides with the phase 2 re-entry phenomenon constituting the repolarization hypothesis in BrS1 78 . The high I Kto density, especially in the right epicardium, results in complete action potential failure 81 . E1784K channels are less available for activation due to the hyperpolarized SSFI-V 1/2 . This seems to be the main mechanism behind the decrease in AP upstroke velocity in cardiac cells, especially the epicardial cells, despite the mutant and triggers-exacerbated increased late I Na . Thus, E1784K expresses both gain-and loss-of-function at the electrical level in cardiac cells. However, this expressivity is finely tuned by channel switches, like temperature and cytosolic calcium.
Our previous and current data suggest exercise, and its accompanying physiological triggers, differentially affect mixed syndrome mutations, especially E1784K 22,37,41,42 . The action of different antiarrhythmics appear to differ depending on physiological state.

Conclusions
Appropriate management of cardiac arrhythmias in SCN5a patients requires careful investigation of antiarrhythmic drug efficacy under various physiological states. Our results suggest that Ranolazine may increase the susceptibility for arrhythmia development in E1784K carriers at sinus rhythm and tachycardia. The risk is augmented under febrile conditions. Although exercise is commonly associated with high heart rates, other pathophysiological states share common triggers, as in heart failure or myocardial ischemia and infarction. Other antiarrhythmics should also be screened against E1784K and other channel mutants under various physiological conditions.

Methods
Homology Modelling and Auto-Docking. Homology modeling was performed using the Swiss-Model server (https://swissmodel.expasy.org) 82 . The newly cryo-EM solved American cockroach voltage-gated sodium channel (Na V Pas) structure (3.8-Å resolution) was used as a template against the Na V 1.5 sequence. Modeling was done according to the protocol established by Bordoli et al. 82 . Sequence alignment was performed using Uniprot Align (http://www.uniprot.org/align/) for SCN5A_HUMAN (Na V 1.5) and SCNA1_PERAM (Na V Pas). Ranolazine was virtually docked using AutoDock4 against the Na V 1.5 homology model built on Na V Pas (Na V 1.5-Na V Pas) 83 . PyMOL-pdb viewer was used for optimization and visualization of the auto-docking results. Cell Culture. HEK293 cells were grown at pH 7.4 in a DMEM (1x) nutrient medium (Life Technologies, NY, USA), supplemented with 10% FBS and maintained in a humidified environment at 37 °C with 5% CO 2 . The α subunits (WT or E1784K) were co-transfected with the β1 subunit and green fluorescent protein, eGFP (1.50 µg: 0.75 µg: 1.50 µg, respectively). The cDNA mixture was then allowed to incubate with the HEK293 cells before plating on coverslips. The HEK293 cells were selected for this study since they contain a relatively elevated [CaM] free level compared to other cell lines, thereby controlling for calcium-calmodulin effects on Na V 1.5 84 . Electrophysiology. Whole-cell patch clamp recordings were performed in extracellular solution containing (mM): 96 NaCl, 4 KCl, 2 CaCl 2 , 1 MgCl 2 , and 10 HEPES (pH 7.4). Solutions were titrated with CsOH to pH 7.4. Pipettes were fabricated with a P-1000 puller using borosilicate glass (Sutter Instruments, CA, USA), dipped in dental wax to reduce capacitance, then thermally polished to a resistance of 1.0-1.5 MΩ. Low resistance electrodes were used to minimize series resistance between pipette and intracellular solution resulting in typical access resistances of 3.5 MΩ or less, thereby minimizing voltage measurement error. Pipettes were filled with intracellular solution. For minimal cytosolic calcium levels, pipettes contained (mM): 130 CsF, 9.6 NaCl, 10 HEPES, and 10 EGTA titrated to pH 7.4. The intracellular pipette solution was manipulated to mimic peak systolic cytosolic calcium 85,86 . To do so, we calculated, using the Ca-EGTA Calculator v1.3, the amount of CaCl 2 (in mM) added to bring cytosolic calcium to 2500 nM at both 22 °C and 34 °C: 9.53 and 9.60, respectively.
All recordings were made using an EPC-9 patch-clamp amplifier (HEKA Elektronik, Lambrecht, Germany) digitized at 20 kHz using an ITC-16 interface (HEKA Elektronik, Lambrecht, Germany). Data were acquired and low-pass-filtered (5 kHz) using PatchMaster/FitMaster software (HEKA Elektronik, Lambrecht, Germany) running on an Apple iMac (Apple Computer, Cupertino, CA). Leak subtraction was performed online using a P/4 procedure. Bath solution temperature was controlled using a Peltier device driven by a TC-10 Temperature Controller (Dagan, Minneapolis, MN). Bath temperature was maintained at 22 °C or 34 °C. Experiments were not performed at physiological temperatures because of the inherent instability of cells at temperatures above 34 °C. We extrapolated data to physiological temperatures using a Q 10 relationship, which was supported with Arrhenius calculations, (described below). After a giga ohm seal resistance was achieved, the whole-cell configuration was attained. The holding potential between protocols was −110 mV. We recorded I Na from cells that expressed currents no greater than −5 nA. The average voltage error calculated for all cells used in this study (n = 250) is 6.06 mV ± 0.40 mV obtained (Table 7). There are no differences between the voltage-errors in the different conditions (p > 0.05). Analysis and Statistics. Analysis and graphing were done using FitMaster software (HEKA Elektronik, Lambrecht, Germany) and Igor Pro (Wavemetrics, Lake Oswego, OR, USA) with statistical information derived using JMP statistical software. Statistical significance was accepted at p < 0.05 using a four-factor completely randomized design (CRD) ANOVA test followed by a post-hoc Tukey test. Our statistical model was a full factorial in Voltage Protocols. Current Density. We measured current density from the ratio of current amplitude to the cell membrane capacitance (pA/pF).

Drug Preparation.
Conductance Density. Channel conductance was calculated from peak I Na using Ohm's law at 0 mV.
where G Na is sodium channel conductance, I Na is peak sodium current in response to the command potential V = 0 mV, and E rev is the reversal potential. We measured conductance density from the ratio of conductance to the cell membrane capacitance (nS/pF).

Activation (GV).
To determine the voltage dependence of activation, we measured the peak current amplitude at test pulse potentials ranging from −100 mV to +80 mV in increments of +10 mV for 19 ms. Prior to the test pulse, channels were allowed to recover from fast inactivation at −130 mV for 197 ms. Channel conductance was calculated from peak I Na using Formula (1). Calculated values for conductance were normalized to the maximal conductance and fit with the Boltzmann function: where G/G max is the normalized conductance amplitude, V m is the command potential, z is the apparent valence, e 0 is the elementary charge, V 1/2 is the midpoint voltage, k is the Boltzmann constant, and T is temperature in °K.

Steady-State Fast Inactivation (SSFI).
The voltage-dependence of SSFI was measured by preconditioning the channels to a hyperpolarizing potential of −130 mV and then eliciting prepulses from −130 or −150 to +10 mV in increments of 10 mV for 500 ms. Channel availability was assessed during a test pulse to 0 mV. Normalized current amplitude as a function of voltage was fit using the Boltzmann function: where I/I max is the normalized current amplitude, z is apparent valence, e 0 is the elementary charge, V m is the prepulse potential, V 1/2 is the midpoint voltage of SSFI, k is the Boltzmann constant, and T is temperature in °K. where I is current amplitude, I ss is the plateau amplitude, α is the amplitude at time 0 for time constant τ, and t is time.
Late I Na Current. Late I Na was measured between 40-50 ms during a 50 ms depolarizing pulse to 0 mV from a holding potential of −130 mV. An average of 10 pulses was used to increase the signal-to-noise ratio.
Use-Dependent Inactivation (UDI, 1 Hz and 3 Hz). Channels accumulated into a use-dependent inactivated state during either a series of 300 380 ms depolarizing pulses to 0 mV followed by a 615 ms-110 mV recovery pulse at a frequency 1 Hz, or 500 220 ms depolarizing pulses to 0 mV followed by a 110 ms-110 mV recovery pulse at a frequency 3 Hz. Normalized current amplitude as a function of time was fit with a double exponential.
where I is current amplitude, I ss is the plateau amplitude, α 1 and α 2 are the amplitudes at time 0 for time constants τ 1 and τ 2 , and t is time.
Q 10 Coefficients. The temperature coefficient for kinetic and thermodynamic parameters plotted as a function temperature was calculated in Igor: where R is the rate and T is temperature (1 and 2 are the two temperatures measured). Rate was calculated by the inverse of the τ value. Q 10 fits for steady-state midpoints and slopes were calculated by replacing the R X with V 1/2 and z values. Fits for y 0 were calculated based of the 1/y 0 to yield optimal Q 10 values. The fit was extrapolated to physiological (37 °C) and febrile (41 °C) temperatures.
Arrhenius Calculations. The Arrhenius linear relationship for the natural exponent of kinetic or thermodynamic parameters as a function of inverse temperature was calculated in Igor:  where k is the rate constant, steady-state midpoint, or slope, A is the pre-exponential factor, Ea is the activation energy, R is the universal gas constant, and T is temperature in °K.
Myocardial Action Potential (AP) Modeling. Simulations. Action potentials were simulated using a modified version of the O'Hara Rudy (ORd) model at 37 °C and 41 °C programmed in Matlab 87 . The sodium data was extrapolated to physiological and febrile temperatures Q 10 values for WT and E1784K at 0 µM and 10 µM Ranolazine. The maximal G Na density was 150 mS/µF in all conditions simulated. We modified the gating I Na parameters data in accordance with our biophysical data for the various conditions. The GV and SSFI midpoints and slopes were extrapolated to 37 °C and 41 °C and normalized to the original ORd parameters. The phosphorylated steady-state fast inactivation midpoints in all channel variants were equally hyperpolarized by 6.2 mV. Late I Na density was normalized to the original ORd value and multiplied by the percentage of late to peak I Na calculated above.  To model the calcium-dependence of our late I Na data, we fit the biophysical parameters extrapolated to 37 °C and 41 °C with a Hill equation: where Z is the biophysical parameter of interest, Y 0 is the minimum value, Y M is the maximum value, X 1/2 is the midpoint of the curve, X is the intracellular cytosolic calcium, b is the rate. Subspace calcium was not accounted for due to the lack of experimental data. Thus, the modified ORd model is a dynamic simulation of the calcium-induced shifts which are observed with increasing intracellular calcium levels as a function of pacing frequency, comprising the positive staircase phenomenon 88,89 .
Simulations at febrile temperature ( , in the ORd model based on previously published Q 10 values. Simulations were run on endocardial, midmyocardial, and epicardial ventricular myocytes using a 0.5 ms stimulus pulse with an amplitude of −80 µA/µF. The stimulus protocol was designed to step up the frequency gradually from 0.5 Hz to 2.5 Hz, with a 1000 beats per frequency step to ensure attainment of steady-state. Analysis. Analysis of APs only included those that fully recovered and were restored to baseline. Action potential duration (APD) was measured at 90% of repolarization by multiplying the resting membrane potential (RMP) value, prior to the current stimulus pulse, by 0.9. The APD 90 of the final two beats in the frequency step were plotted versus the diastolic interval (DI = BCL − APD 90 ), where BCL is the basic cycle length, creating electrical restitution curves.
Data Availability. All data generated or analysed during this study are included in this published article.