Flufenamic acid decreases neuronal excitability through modulation of voltage-gated sodium channel gating

Abstract

The electrophysiological phenotype of individual neurons critically depends on the biophysical properties of the voltage-gated channels they express. Differences in sodium channel gating are instrumental in determining the different firing phenotypes of pyramidal cells and interneurons; moreover, sodium channel modulation represents an important mechanism of action for many widely used CNS drugs. Flufenamic acid (FFA) is a non-steroidal anti-inflammatory drug that has been long used as a blocker of calcium-dependent cationic conductances. Here we show that FFA inhibits voltage-gated sodium currents in hippocampal pyramidal neurons; this effect is dose-dependent with IC₅₀ = 189 μm. We used whole-cell and nucleated patch recordings to investigate the mechanisms of FFA modulation of TTX-sensitive voltage-gated sodium current. Our data show that flufenamic acid slows down the inactivation process of the sodium current, while shifting the inactivation curve ∼10 mV toward more hyperpolarized potentials. The recovery from inactivation is also affected in a voltage-dependent way, resulting in slower recovery at hyperpolarized potentials. Recordings from acute slices demonstrate that FFA reduces repetitive- and abolishes burst-firing in CA1 pyramidal neurons. A computational model based on our data was employed to better understand the mechanisms of FFA action. Simulation data support the idea that FFA acts via a novel mechanism by reducing the voltage dependence of the sodium channel fast inactivation rates. These effects of FFA suggest that it may be an effective anti-epileptic drug.

Introduction

Action potential generation is crucial for nervous system function. Fast neuronal action potentials are mainly shaped by voltage-gated sodium and potassium channels (Hodgkin & Huxley, 1952). The gating properties of tetrodotoxin (TTX)-sensitive, voltage-gated sodium channels play a fundamental role in determining the diversity in intrinsic excitability and firing patterns found among different types of neurons (Martina & Jonas, 1997; Raman & Bean, 1997). The functional properties of sodium channels depend on both the expression of different α channel subunits (Raman & Bean, 1997; Chen et al. 2008; Lee & Goldin, 2008) and the presence of various β subunits, which strongly influence the onset of inactivation (Isom et al. 1992; Makita et al. 1996) and modulate the persistent and resurgent sodium currents (Grieco et al. 2005; Aman et al. 2009). Additionally, the metabolic state of the neurons also affects sodium current properties via G-protein- and kinase-dependent mechanisms (Ma et al. 1994; Cantrell et al. 1996).

The processes of voltage-dependent inactivation and recovery from inactivation determine the sodium channel availability both at resting membrane potential in a quiescent neuron and following the firing of one or more spikes. Therefore, sodium channel inactivation properties are critical for reliable transmission of action potential trains (Jung et al. 1997; Martina & Jonas, 1997) and for generation of burst firing (Azouz et al. 1996; Swensen & Bean, 2003). Thus, pharmacological modification of the inactivation properties can dramatically affect the ability of neurons to fire action potential in trains and/or bursts.

Flufenamic acid (FFA), a non-steroidal anti-inflammatory drug, has effects on multiple types of ion channels. For years this drug has been used as a blocker of calcium-activated cationic currents (Shaw et al. 1995; Peña et al. 2004) and more recently it has been shown to reduce calcium-dependent hyperpolarization in lamprey spinal cord, probably through an inhibition of calcium currents (Wang et al. 2006) and to potentiate leakage potassium currents mediated by a subset of KCNK channels, namely TREK and TRAAK (Takahira et al. 2005). Here we show that FFA has an interesting modulatory effect on neuronal sodium channels, reducing sodium current availability and slowing down inactivation and recovery from inactivation, leading to diminished repetitive and burst firing. Computer simulations indicate that these effects are the consequence of modified transition rates between the open and the fast inactivated state.

Methods

Hippocampal brain slices

All experiments followed protocols approved by the Northwestern University Center for Comparative Medicine and complied with the policies and regulations described by Drummond (2009). Ten- to twenty-day-old Long–Evans rats were anaesthetized with isoflurane and killed by decapitation. The brain was removed from the skull in ice-cold artificial cerebrospinal fluid (ACSF), containing (in mm): 125 NaCl, 25 NaHCO₃, 2.5 KCl, 1.25 NaH₂PO₄, 1.8 CaCl₂, 1 MgCl₂ and 25 glucose, bubbled with 95% O₂ and 5% CO₂ (pH 7.4). Transverse hippocampal slices (300 μm thick) were cut using a vibroslicer (Dosaka) and stored in a solution containing (in mm): 87 NaCl, 25 NaHCO₃, 2.5 KCl, 1.25 NaH₂PO₄, 0.5 CaCl₂, 7 MgCl₂, 75 sucrose and 25 glucose, bubbled with 95% O₂ and 5% CO₂; slices were kept at 35°C for 15–20 min and subsequently at 22–24°C.

Electrophysiological recordings

Slices were visualized with an Axioskop 2FS (Zeiss) upright microscope with a water-immersion 60× objective (0.9 NA, Olympus). For whole-cell current-clamp recordings, the bath solution contained kynurenic acid (2 mm) and picrotoxin (0.1 mm) to block fast synaptic transmission, and pipettes were filled with internal solution consisting of (in mm): potassium gluconate 140, NaCl 8, MgCl₂ 2, EGTA 1, Na₂ATP 2, NaGTP 0.1, Hepes 10, pH 7.3 with KOH. All voltage-clamp recordings were performed at 24–25°C, while current clamp recordings were performed at 32–33°C.

In order to study the gating properties of the sodium channels, voltage-clamp recordings were performed in the nucleated patch configuration, which allows almost ideal voltage clamp, even for fast sodium currents (Martina & Jonas, 1997). Recordings were performed at 24–25°C using an Axopatch 200B (Axon Instruments) patch-clamp amplifier. Data were sampled at 20 kHz and filtered at 10 kHz. Patch pipettes had a resistance of 2.5–4 MΩ (in working solution) and were pulled from 1.65 mm OD, 1.1 mm ID, WPI glass (PG10165-4). The internal solution was CsCl-based and contained (in mm): CsCl 140, NaCl 10, MgCl₂ 2, EGTA 10, Na₂ATP 2, NaGTP 0.1, Hepes 10, pH 7.3 with CsOH.

Capacitive transients were reduced by wrapping the pipettes in parafilm and compensated using the fast compensation circuitry of the amplifier. To obtain nucleated patches, the whole-cell configuration was obtained and the pyramidal identity of cells was quickly confirmed by the presence of a sag in the membrane potential in response to a hyperpolarizing current injection. The pipette was then slowly withdrawn while applying a constant negative pressure (−90 to −130 mbar). After excision, patches were held at −90 mV with a small pressure (−0.5 to −1.5 KPa). All recordings (except for Fig. 1 that was obtained using a P/–5 protocol) were performed first in control conditions and then in the presence of TTX (500 nm). The sodium currents were then obtained by offline digital subtraction of traces in TTX from control traces.

Figure 1. Flufenamic acid reduces voltage-gated sodium current in hippocampal pyramidal neurons.

A, Na⁺ currents recorded from the same nucleated patch in control condition (left), in the presence of 200 μm FFA (middle) and after drug wash-out (right); a 50 ms pre-pulse to −75 mV preceded the 30 ms test pulse to −10 mV (bottom); patches were held at −90 mV to avoid slow inactivation. FFA (200 μm) blocked ∼50% of the current. B, dose–response curve for FFA on currents elicited by a test 30 ms pulse to −10 mV from a holding of −75 mV. Fitting the data with a Hill equation returned an IC₅₀ value of 189 μm and a Hill coefficient of 2.5.

Drugs were applied to the patches by using a multi-barrel system consisting of four glass capillaries (1.1 mm ID) glued together and connected to a syringe pump (WPI). The patch-pipette carrying the nucleated patch was sequentially inserted into the barrels, each containing a different extracellular solution. This system ensures that the solution speed is the same in each barrel. One limiting factor is that completion of pipette placement requires ∼10 s; during this time the patch experiences diverse and unsteady drug concentrations. Thus, this is the limiting step for measuring the time course of drug action. In experiments testing the effect of FFA, after the patch-pipette was inserted in the FFA barrel, the current was constant and no build-up effect was detected. Thus, the onset of FFA effect is faster than the time resolution of our experimental protocol. We can say that after 10 s of application the effect had reached steady state. Thus, FFA is a relatively rapidly acting drug.

For these recordings, the extracellular solution consisted of Hepes-buffered ACSF (in mm): 138 NaCl, 10 Hepes, 2.5 KCl, 2 CaCl₂, 1 MgCl₂ and 25 glucose (pH 7.3 with NaOH), 50 μm CdCl₂, 10 mm TEA plus the drugs of interest.

All drugs were from Sigma, except TTX (Alomone). FFA was prepared as stock solution (200 mm in DMSO) and stored at 2–4°C. Working solutions were prepared daily. The control solutions always contained DMSO at the same concentration as in the test solution. 4-Aminopyridine (4-AP) was prepared as stock solution (100 mm in Hepes-buffered ACSF) and stored at −30°C. TTX was prepared as stock solution (1 mm in H₂O) and stored at 2–4°C.

Data acquisition and analysis

Data were transferred to a PC using a Digidata 1322A (Axon Instruments) interface and acquired using pCLAMP 9 (Axon Instruments) software. Activation curves were obtained from the peak current–voltage relation using the Goldman–Hodgkin–Katz current equation (Hille, 2001):

where I_Na represents the measured sodium current, P_Na the permeability and z the ionic charge (1 for Na⁺ ions), E the membrane potential, and T, F and R, the absolute temperature, Faraday's and gas constants, respectively. The measured peak current values were transformed into permeability, normalized to their maximum value, plotted against membrane potential and fitted with a Boltzmann function raised to the third power of the form: f(V) = 1/(1 + exp(−(V−V_1/2)/k))³, where V_1/2 is the voltage at which the activation is half-maximal, and k is the slope factor. The voltage dependence of inactivation was fitted using a simple Boltzmann function, f(V) = 1/(1 + exp ((V−V_1/2)/k)). The time course of the inactivation onset could always be fitted with a single-exponential function. This was supported by the fact that the traces could be fitted by a straight line in semi-logarithmic plots (Supplemental Fig. S1). The time course of recovery from inactivation was fitted with a bi-exponential equation of the form, f(t) = 1 − (A1 exp(−t/τ₁) +A2 exp(−t/τ₂)), yielding two time constants (τ) and relative contributions (A). A1 + A2 was constrained to 1.

The dose–response curve was obtained by fitting the experimental data with a logistic equation with variable slope (Hill coefficient) where: Fraction inhibited = C/(1 + (IC₅₀/[FFA])ⁿ), where IC₅₀ represents the half-maximal concentration, n represents the Hill coefficient and C represents the asymptotic value of the function.

Action potential threshold values were obtained from the current-clamp whole-cell recordings and calculated using two methods (that produced the same results): in one the threshold potential was identified as the inflection point in the trace of the second derivative of the membrane potential (plotted versus membrane potential, Kress et al. 2008). In the second method the first derivative of the membrane potential was calculated and the threshold determined as the first point where it exceeded 4 times the standard deviation of the baseline. The recordings for these measurements were sampled at 20 kHz (and filtered at 10 kHz), but traces were interpolated offline using a cubic spline method in order to obtain smoother traces.

The data points shown in the figures and the mean values given in the text were obtained by pooling data from different patches. Individual experiments were also analysed separately to test statistical significance and to obtain standard errors. All results are reported as mean ± standard error. Statistical significance was assessed using Student's t test (paired or unpaired, two-tailed) at the significance level P < 0.05.

Computational modelling

Simulations were performed on a dual core 1.83 GHz Inspiron 640 PC using the fully implicit backward Euler integration method with a time step of 0.01 ms in the NEURON simulation environment (http://www.neuron.yale.edu/neuron/, version 5.8, (Hines, 1993; Carnevale & Hines, 2006)). We used a modified version of a previously published six-state sodium current model (Baranauskas & Martina, 2006). In this modified version, a slow inactivated state has been added in order to account for the presence of a slow phase in the recovery from inactivation and for the differences in the voltage dependence of the steady-state inactivation measured using 50 ms and 200 ms pre-pulses. The experimental data show that FFA had only a minor effect on this slow inactivation process and, accordingly, the transition rates to the slow inactivated state are not affected by FFA in the model. In addition, in the modified model the voltage dependence of the activation rates was tuned to match the more depolarized half-activation voltage and a shallower voltage dependence of activation. The transition rates between different states were described by a general equation which helps to relate the transition rates to the changes in the Gibbs free energy W and the moving particle charge (Hille, 2001): α/β = 1/(A+ (1/r)*exp[−(W−z_α/β*v)/(kT/q_e)]).

Here α and β correspond to the forward and backward transition rates as shown in Fig. 8; A accounts for the rate saturation; 1/r corresponds to the inverted Kramers escape rate, W is the Gibbs energy associated with the voltage-independent conformational change; z_α/β is the product of the voltage sensor charge and the fraction of the membrane voltage that the particle senses during gating movements; k is the Boltzmann constant; T is the absolute temperature and q_e is the electron charge. The values of these parameters are provided in Table 1. It was assumed that FFA only affects the transition to the fast-inactivated state by modifying the corresponding z_α/β and W values. Moreover, it was assumed that the ratio of gating charges for this transition, z_α/z_β, was not affected by FFA. Thus, modifications in only two parameters were introduced and sufficient to simulate FFA effects.

Figure 8. Flufenamic acid decreases 4-AP-induced bursting in CA1 pyramidal neurons.

A, current clamp recordings obtained from a CA1 hippocampal neuron in response to depolarizing current (120 pA, 1 s) injection in the presence of 4-AP (100 μm, left panel) and 4AP+FFA (100 μm, right panel). The trace segments between dashed lines are shown on an expanded time scale (insets) to resolve bursts. B, firing frequency in response to depolarizing current injections. C, interspike interval ratio (ISI) against different current injections. For each current step, the ISI ratio was obtained dividing the interval between the first two spikes by the mean interspike interval. Control: n = 7; FFA: n = 7. Recordings were obtained at 32–33°C.

Table 1.

Rate constants

	q10	A	1/r	W (mV)	zα,β
α1	2.8	0	0.13	−36.3	0.57
β1	2.8	0	0.13	−36.3	3.2
α2	2.8	0.036	0.58	−67.2	1.28
β2	2.8	0.043	0.058	−67.2	1.28
α3	2.4	0.3	8	−102	2.2
β3	2.4	0	8	−102	1.47
α3 in FFA	2.4	0.3	9	−80	1.4
β3 in FFA	2.4	0	9	−80	0.93
α4	2.4	0.3	1100	−198	2.7
β4	2.4	0	1100	−198	2.8

The general equation for rate constants is:

α,β = q10([23–T]/10)/(A+ (1/r) * exp[–,+(W–z_α×V)/(kT/q_e)])

Here kT/q_e is equal to 25.6 mV for 24°C.

Voltage-clamp recordings were simulated assuming [Na⁺]_i = 5 mm and [Na⁺]_o = 140 mm.

Results

The effect of FFA on sodium currents and cell firing of CA1 pyramidal neurons was studied by performing recordings in acute slices obtained from 11- to 20-day-old rats.

In a first set of experiments, nucleated patch recordings were employed to investigate the effect of FFA on the amplitude of sodium currents recorded in nucleated patches using a 30 ms step (to 10 mV) from −75 mV (close to the resting potential of hippocampal pyramidal neurons). In these first measurements FFA was tested at 0.2 mm, a concentration similar to that used as the calcium-activated non-specific cationic current (ICAN) blocker (0.5 mm, Peña et al. 2004), as a two-pore potassium channel opener (0.1 mm, Takahira et al. 2005) and as blocker of calcium-dependent afterhyperpolarization (0.2 mm, Wang et al. 2006).

We found that 200 μm FFA blocked ∼50% of the peak sodium current elicited from a holding potential of −75 mV; this block was promptly reversible upon washout of the drug (Fig. 1A). Having discovered a robust effect of FFA at this concentration, we proceeded to obtain a dose–response curve for the Na⁺ current block by FFA. Fitting the experimental data with a logistic equation returned an IC₅₀ value of 189 μm, a slope factor of 2.5 and a maximum efficacy of 86% (Fig. 1B). All the remaining experiments were performed using FFA at 200 μm. Two reasons prevented us from using higher concentrations: first, with higher concentrations the current size in the patches is too small to allow reliable measurements and, second, at higher FFA concentrations patches are quite short-lived and TTX subtractions become extremely difficult.

Next, we characterized in more detail the mechanism of action of FFA (0.2 mm) on sodium channels. Even with a caesium-based intracellular solution and in the presence of extracellular TEA and cadmium, some contaminating currents could often still be detected in pyramidal cell nucleated patches; therefore our analysis was performed on TTX-subtracted traces only. Consequently, it was not possible to study the TTX-sensitive current in the same patch before and after FFA. Thus, for all the experiments described next, the control and the FFA data were obtained from different patches.

FFA had only minimal effect on the Na⁺ channel activation curve in hippocampal pyramidal cells (Fig. 2). Permeability and voltage plots showed that the activation curve was slightly affected by FFA (mid-point and slope factor were −12.5 ± 1.78 mV and 4.5 ± 0.25 mV in control versus−11.3 ± 1.36 mV and 5.6 ± 0.27 mV in FFA, respectively, Fig. 2C). In contrast, the inactivation process was strongly affected by FFA. Inactivation was fitted with a single exponential function (Supplemental Fig. S1); in the presence of FFA, inactivation was significantly slower than in the control condition at most membrane potentials (Fig. 3; at 0 mV, the time constant was 0.48 ± 0.03 ms in control (n = 15) and 0.70 ± 0.03 ms in FFA (n = 16), P < 0.001). Thus, the mode of action of FFA appears to differ from typical use-dependent blockers such as carbamazepine or imipramine, which induce faster inactivation (Yang & Kuo, 2002). We then analysed the effect of FFA on the voltage-dependent channel availability. For these experiments, patches were held at −90 mV, and tested using either a 50 ms or a 200 ms pre-pulse from −120 to −30 mV (10 mV steps) preceding a 30 ms test pulse to −10 mV (Fig. 4, insets). FFA dramatically shifted the steady-state inactivation curve to more hyperpolarized potentials; this shift explains the large reduction of the current amplitude at resting membrane potentials. Interestingly, the shift was similar for relatively short (50 ms, Fig. 4A and B) and for longer pre-pulses (200 ms, Fig. 4C and D). For a 50 ms pre-pulse, the mid-point potential shifted from −55.4 ± 1.3 mV in control (n = 13) to −67.8 ± 1.4 mV in FFA (n = 16), and the slope factor from 8.3 ± 0.4 to 10.7 ± 0.5 mV. For a 200 ms pre-pulse, the mid-point potential shifted from −64.1 ± 2.2 mV in control (n = 8) to −74.1 ± 2.6 mV in FFA (n = 4) and the slope factor from 7.5 ± 0.2 to 8.9 ± 1.0 mV.

Figure 2. Flufenamic acid has minimal effect on Na⁺ channel activation.

A, representative traces of Na⁺ currents recorded from nucleated patches in control conditions (top traces) and in the presence of 200 μm FFA at the potential indicated in the figure. The bottom traces depict the stimulation protocol for these traces. B, I–V curve from 15 patches recorded in control conditions (open symbols) and 16 in the presence of 0.2 mm FFA (filled symbols). Patches were held at −90 mV, stepped to −120 mV for 50 ms, then depolarized by a family of 30-ms-long pulses between −70 mV and +40 mV (10 mV steps), and back to −90 mV. C, Na⁺ permeability, calculated from the peak current (see Methods) in control (open symbols) and in the presence of FFA (filled symbols), normalized to the maximum value and plotted against test pulse potential. The experimental data were fitted with Boltzmann functions raised to third power. No significant difference was found in mid-point potential between control (−12.5 ± 1.8 mV, n = 15) and FFA treatment (−11.3 ± 1.36 mV, n = 16) while the value of the slope factor (k/3) in FFA (5.6 ± 0.27 mV) was slightly larger than in control (4.5 ± 0.25 mV, P < 0.01).

Figure 3. Flufenamic acid slows Na⁺ channel inactivation.

A, the time course of Na⁺ channel inactivation at 10 mV (30 ms pulse) in control and in the presence of FFA. Na⁺ currents were recorded using test pulses between −30 and 40 mV in 10 mV increments following a 50-ms-long pre-pulse to −120 mV. The traces shown were normalized to their respective peak current amplitude and superimposed to better illustrate the effect of FFA on the current inactivation and fitted using single exponential curves (dashed lines). B, Na⁺ current inactivation time constant at different potentials in control (n = 15, open symbols) and in FFA (n = 16, filled symbols). The inactivation time course could be fitted with a single exponential function at all membrane potentials tested.

Figure 4. Flufenamic acid shifts Na⁺ current steady-state inactivation and decreases Na⁺ availability at resting membrane potential.

A and C, Na⁺ currents recorded in control and in the presence of FFA. Nucleated patches were held at −90 mV, and currents were elicited by 30 ms test pulses to −10 mV preceded by either 50 ms (A) or 200 ms (C) pre-pulses (from −120 to −30 mV, 10 mV steps, insets). B and D, steady state inactivation curves with 50 ms (B) or 200 ms (D) pre-pulses in control (open symbols) and in FFA (filled symbols). Peak Na⁺ current amplitudes were normalized to the maximum current, plotted against pre-pulse potential value, and fitted by simple Boltzmann equations. For 50 ms pre-pulse, the mid-point potential and slope factor were −55.4 ± 1.3 mV and 8.3 ± 0.4 mV in control (n = 13), and −67.8 ± 1.4 mV and 10.7 ± 0.5 mV in FFA (n = 16). For 200 ms pre-pulse, the mid-point potential and slope factor were −64.1 ± 2.2 mV and 7.5 ± 0.2 mV in control (n = 8), and −74.1 ± 2.6 mV and 8.9 ± 1.0 mV in FFA (n = 4), respectively.

A large shift in the inactivation curve suggests that differences may exist also in the process of the recovery from inactivation. To investigate this hypothesis, standard double-pulse protocols, in which two identical voltage pulses (to −10 mV, 30 ms) are separated by an interval of increasing duration (Fig. 5A and B), were used to study the recovery from inactivation. We focused on the recovery in the millisecond range and we did not investigate very slow inactivation that happens over seconds (Fleidervish et al. 1996) and that may involve different mechanisms. Because only this faster recovery process, occurring in the millisecond timescale, was investigated, here we refer to ‘fast’ and ‘slow’ inactivation only to make a distinction between the two inactivated states characterized by faster and slower recovery, although even the slower recovery occurs, at −120 mV, with a time constant of <200 ms. A first set of experiments was performed holding the patches at −90 mV (with 50 ms pre-pulses to −120 mV, see Fig. 5A). As described previously (Martina & Jonas, 1997), recovery from inactivation of the sodium channels of CA1 pyramidal neurons was best described by a bi-exponential function and this was the case in the presence of FFA as well (Fig. 5C). Intriguingly, the fast and slow time constants were differentially affected by FFA; while the fast time constant was largely increased (∼3-fold, from 1.7 ± 0.3 ms in control to 4.9 ± 0.7 ms in FFA; n = 6–7 and 5–9, respectively, P < 0.05), no appreciable effect was detected on the value of the slow component (65 ± 39 ms in control versus 51 ± 91 ms in FFA, Fig. 5C). Indeed, no significant FFA effect was observed for any interval >20 ms. In a second set of recordings, recovery from inactivation was tested at −70 mV, a more physiologically relevant membrane potential. Interestingly, no appreciable effect on the recovery from inactivation was detected at this holding potential (Fig. 5D), showing a voltage-dependent effect in agreement with that on inactivation rate (Fig. 3). These experiments showed that FFA affects the speed of the repriming process in a voltage-dependent manner.

Figure 5. Flufenamic acid slows the recovery of Na⁺ channel from inactivation.

A and B, recovery from fast inactivation of Na⁺ currents in control (A) and in the presence of FFA (B). A double-pulse protocol was used to measure the extent of recovery from inactivation. Patches were held at −90 mV, and Na⁺ currents were elicited by a 30 ms test pulse to −10 mV (following a 50 ms pre-pulse to −120 mV); a second identical test pulse at −10 mV was then applied after increasing time intervals at −120 mV. These traces were obtained using interpulse intervals from 1 to 12 ms and normalized to the peak current elicited by the respective first pulses for better comparison. C, time course of recovery from inactivation in control and in FFA at −120 mV. The amplitude of Na⁺ current evoked by the second test pulse was normalized to that of the first test pulse and plotted against the interpulse interval. Data points were fitted using double exponential functions. The fast and slow components of recovery time constants for the pooled data were 1.7 ms (A = 0.77) and 65 ms in control (n = 6–7), and 4.9 ms (A = 0.76) and 51 ms in FFA (n = 5–9). D, time course of recovery from inactivation in control and in FFA at −70 mV. Fitting the pooled data with a double exponential function returned values of the fast and slow time constants of 12.5 ms and 787 ms in control and 12.3 and 674 ms in FFA. The slow component contribution was 34% and 45% in control and FFA, respectively.

The voltage-clamp data show that FFA modulates Na⁺ channel availability and recovery from inactivation, which are critical for repetitive and burst firing. Therefore, we tested the effect of FFA on burst and tonic firing of hippocampal pyramidal neurons (Fig. 6). All these recordings were obtained in the presence of kynurenic acid (2 mm) and picrotoxin (0.1 mm) to block fast synaptic transmission. Under these conditions FFA (0.2 mm) did not produce any significant effect on resting membrane potential (−68.4 ± 0.8 mV in control versus−66.9 ± 0.7 mV in FFA, n = 8, Supplemental Fig. S2) and only a small decrease in input resistance (from 177 ± 12 MΩ in control to 150 ± 8 MΩ in FFA, P < 0.05, n = 23, paired t test; Supplemental Fig. S2), FFA, however strongly decreased the number of action potentials elicited by 1 s depolarizing current injections and the instantaneous firing frequency. For a 150 pA depolarizing current pulse the average number of spikes elicited in 1 s decreased from 20.9 ± 2 in control to 1.9 ± 0.6 in FFA, n = 23 and 31, respectively, Fig. 6B), while the instantaneous firing frequency decreased from 23.9 ± 2.1 Hz in control conditions to 17.4 ± 2.6 Hz in FFA (Fig. 6A and C). Rheobase current increased about 2-fold, from 76 ± 9 pA to 163 ± 11 pA (P < 0.05). This effect on firing phenotype was accompanied by small, but significant effects on the action potential threshold, which changed from −44.3 ± 0.9 mV in control to −42.2 ± 1.4 mV in FFA (n = 23, P < 0.05, Fig. 7C–E), height (from 99.0 ± 1.2 mV to 95.4 ± 1.4 mV, P < 0.05, Fig. 7B) and half-duration (from 1.26 ± 0.06 ms to 1.44 ± 0.1 ms, P < 0.05, Fig. 7F).

Figure 6. Flufenamic acid decreases pyramidal cell firing upon depolarizing current injection.

A, current clamp recordings obtained (at 32–33°C) from a CA1 hippocampal neuron in response to depolarizing current injection (150 pA, 1 s, bottom) in control (top trace) and in the presence of FFA (200 μm, middle). B and C, total number of spikes (B) elicited in the cells during 1 s current injection and, (C) instantaneous firing frequency in response to a series of current injections (1 s, 30 pA steps) in control (n = 31) and in FFA (n = 23).

Figure 7. Effects of flufenamic acid on individual action potentials of pyramidal cells.

A, action potentials elicited by near-rheobase current injections in an individual neuron in control conditions and in the presence of FFA (0.2 mm). B, bar chart showing the action potential height. The values were 99.0 ± 1.2 mV and 95.4 ± 1.4 mV in control conditions and FFA, respectively (n = 23, P < 0.05). C, the first derivative of the membrane potential traces shown in A, plotted versus membrane potential. D, bar chart of the value of action potential threshold. E, the second derivative of the membrane potential traces shown in A, plotted versus membrane potential. F, values of action potential half-duration in control and in the presence of FFA (all recordings at 32–33°C).

Next, we examined whether FFA also inhibits burst firing. To do so, we exploited the fact that all CA1 pyramidal cells show robust burst firing in the presence of 4-aminopyridine (4-AP), a potassium channel blocker. Whole-cell current-clamp recordings were obtained from CA1 pyramidal neurons in 4-AP-bathed slices. These recordings were also performed at 32–33°C and in the presence of kynurenic acid and picrotoxin. As expected, 4-AP induced robust burst firing (Fig. 8); bath application of FFA (0.2 mm) dramatically reduced both the total number of spikes and bursting (Fig. 8B and C), as quantified using the interspike interval ratio, which is inversely related to bursting (Metz et al. 2005, 2007). The interspike interval ratio was significantly increased for all current injections larger than 120 pA (for 120 pA, the ISI ratio went from 0.38 ± 0.14 in control to 0.74 ± 0.06 in FFA, P < 0.05, Fig. 7C).

Having described the modulation of neuronal sodium currents by FFA and the functional consequences on neuronal firing, we then used a computational model to obtain further insight into the mechanism of action of FFA. Simulations were also employed to test to which extent FFA effects on sodium channels alone may reproduce the effects observed in current-clamp. We employed a modified version of our previously published model (Baranauskas & Martina, 2006). First, the model was tuned to match the properties of sodium currents in hippocampal neurons such as a slightly different voltage dependence of activation and the presence of a slow phase in the recovery from inactivation (Fig. 9A and C, Table 1, see Methods). To account for the slow phase in the recovery from inactivation, a slow inactivated state was added (Fig. 9A). FFA effect on sodium channels is mostly limited to the fast inactivation (Figs 2–5). Accordingly, in the model the transitions to the fast inactivated state are the only ones affected by FFA. These transitions are described in the framework of the absolute rate theory that relates the transition rates to the Gibbs energy associated with the voltage-independent conformational change during the gating and to the z factor, the product of the gating particle charge and the fraction of the membrane potential sensed by that particle. In our model, the forward and backward transitions between the open and fast inactivated states have different z factors that reflect different voltage dependence of inactivation and recovery from inactivation rates. However, it was assumed that FFA only modifies the sum z_α+z_β, but not the ratio z_a/z_b. In addition, in our model FFA affected the Gibbs energy, W, associated with the voltage-independent conformational change. Thus, all FFA effects were reproduced by a change in two model parameter values that modified the voltage dependence of the fast inactivation rates and corresponding time constants (Fig. 9B and C). Figure 9D shows that the simulated currents are similar to the ones recorded in hippocampal neurons and that FFA changed very little the activation properties of these currents while the inactivation rate was slower in FFA (Fig. 9E). Nevertheless, there was a dramatic change in excitability (Fig. 9F–H); although in control conditions a cell fired regular action potentials in response to current injection (Fig. 9F), in the presence of FFA only few action potentials were fired and then the neuron became quiescent (Fig. 9G). In addition, much larger currents were necessary to evoke action potentials (Fig. 9H).

Figure 9. Computational model of the flufenamic acid effect on sodium current gating.

A, schematics of model states and transitions between them. The upper row corresponds to the slow inactivated states while the bottom row corresponds to the fast inactivated states. The middle row includes two closed states and one open state. Only the open state passes current. The forward transition rates are shown above/on left while the backward rates are shown below/on right of the corresponding transition arrows. B, time constants of inactivation calculated as τ = (α+β)⁻¹. The slow inactivation rates were unaffected by FFA. C, transition rates. D. simulated voltage clamp recordings (protocol shown on top, lighter wide traces) overlaid with traces obtained during recordings from hippocampal neurons (dark thin traces). E, simulated currents evoked during voltage clamp protocol shown on the top. The black trace was obtained with a control model while the grey trace represents the FFA-modified model. F and G, current clamp traces simulated with control (F) and FFA-modified (G) model. Scale bars correspond to 2 ms and 100 pA in C and to 200 ms and 20 mV in E and F. H, summary plot of simulation results for a number of action potentials evoked by a 1 s current pulse in control (filled triangles) and FFA (open circles). The estimated rheobase showed an ∼2-fold increase, from ∼140 pA in control to ∼250 pA in FFA.

These effects are explained by the reduced availability of sodium channels at membrane potentials close to the rest as indicated by summary graphs in Fig. 10. Indeed, in the model FFA shifted inactivation voltage dependence to more hyperpolarized potentials (Fig. 10A and B, filled circles), closely matching the experimental data (Fig. 10A and B, lines). In addition, the recovery from inactivation was slowed down at −120 mV (but not at −70 mV, Fig. 10C–E), also reproducing the experimental data. Thus, the proposed model accounts for most of the FFA effects on sodium current and shows that, although effects on other channels are likely (see below), sodium channel modulation can by itself qualitatively explain many of the observed current-clamp data.

Figure 10. Comparison of the inactivation properties of simulated and experimentally measured currents.

A, comparison of h_∞ curve (obtained with a 50 ms pre-pulse) in experimentally recorded and simulated traces. Experimental data points are represented by continuous fit functions while filled circles represent simulated current measurements. B, similar plots obtained using 200 ms pre-pulses. C, recovery from inactivation at −120 mV. Experimental data are the same as in Figs 4 and 5. D, E, Recovery from inactivation at −70 mV; the faster phase of the recovery (first 120 ms) is shown on an expanded time scale in D. Experimental data are the same as in Fig. 4 and 5.

Discussion

We show that FFA down-modulates TTX-sensitive sodium current in central neurons. At near-resting membrane potential this modulation is mainly the consequence of the large shift in the voltage-dependent inactivation curve, while a slower recovery from inactivation may play a role only at more hyperpolarized membrane potential. Computer simulations indicate that both effects can be accounted for by a change in the voltage sensitivity of the fast inactivation gate. At cellular level, FFA modulation of the sodium current leads to reduction in repetitive firing and elimination of burst firing, two features that suggest an anti-epileptic effect of this molecule.

Mechanism of Na⁺ channel modulation by flufenamic acid

The shift in voltage dependence of inactivation (∼12 mV for a 50 ms pre-pulse and 0.2 mm FFA) leads to a substantial reduction in the fraction of sodium channels available for activation at membrane potentials close to the resting potential in pyramidal neurons (from 0.86 ± 0.02 in control to 0.57 ± 0.04 in FFA, at −70 mV, Fig. 1). However, because of the large safety factor for action potential generation in pyramidal neurons (Madeja, 2000) and because of the negligible effect of FFA on the activation curve of sodium channels, single action potentials were little changed by FFA (Fig. 7), although rheobase showed a 2-fold increase. Repetitive firing, however, was largely reduced and burst firing abolished. This effect is reminiscent of the use-dependent block by lidocaine that has been attributed to a particularly high affinity of lidocaine to the inactivated channel (Hille, 1977; Bean et al. 1983), and is also similar to the effects of numerous anti-epileptic drugs such as phenytoin and carbamazepine (Rogawski & Loscher, 2004). However, the mechanism of FFA action appears to differ from the use-dependent block reported in the literature because the inactivation becomes slower rather than faster and, crucially, there is no change in the slow phase of recovery from inactivation. In hippocampal pyramidal neurons the recovery from inactivation is best described by the sum of two exponential functions (Martina & Jonas 1997; this paper). Use-dependent block dramatically increases the slower phase of recovery from inactivation (Lang et al. 1993) and this leads to larger cumulative inactivation of sodium currents during application of brief repetitive depolarizing pulses (Willow et al. 1985). Such an increase in the slow cumulative inactivation should be present if a drug causes use-dependent inactivation (Hille, 2001; Rogawski & Loscher, 2004), but our data show no change in the slow phase of recovery from inactivation. Thus, an alternate explanation should be found for FFA action and, to this end, we developed a computational model to study the mechanisms of FFA action on sodium channel gating.

Simulations showed that the effects of FFA on sodium currents could be reproduced by assuming a change in both the on and off transition rates between the open and the fast inactivated state. To reproduce FFA effects in the simulations the gating particle charge had to be reduced by ∼30% while the change in voltage-independent conformational energy was <kT, here kT is the product of the Boltzmann constant, k, and the temperature, T. Since, in ideal gases, the average thermal energy of an atom is 1.5 times kT, a reduction in Gibbs energy of less than kT can be considered as marginal. Thus, FFA effects on sodium currents can be interpreted as screening by FFA of the fast inactivation gate voltage sensor. Interestingly, our simulations yielded a value for the z factor that is different for forward and backward transitions. One possible explanation for this result is that, according to Kramers’ theory, the reaction rates are determined solely by the energy difference between the starting point and the barrier energy (Kramers, 1940). In such a model, the voltage dependence of the rate constant is due to an additional energy that the gating particle charge senses while moving from the initial state to the barrier. Thus, it is only the fraction of the electrical field between the initial state and the barrier point that affects the rate and, if a barrier is not at the electrical mid-point of the applied electric field, this fraction will be different for forward and backward movements.

Structurally, FFA is similar to several commonly used anti-epileptic drugs (AEDs), which are characterized by two phenyl groups separated by one or two C–C or C–N single bonds (Rogawski & Loscher, 2004), a trait maintained in the structure of FFA. It has been suggested that the pharmacophoric group in AEDs is defined by a polar group and a hydrophobic group that may be aliphatic or aromatic but should be no smaller than three atoms (Tasso et al. 2001). These features are present in FFA, although the mechanism of action on the Na⁺ current appears different, as it slows down rather than accelerates inactivation; however, while the simulations are consistent with the experimental data they cannot exclude other mechanisms, and more studies, possibly taking advantage of specific mutations in the sodium channel subunits, will be required to address this point.

Flufenamic acid abolishes burst firing

In physiological conditions about 15% of pyramidal neurons in the hippocampal CA1 area are burst firing (Metz et al. 2005). This firing phenotype is regulated by numerous conductances, some of which act synergistically and other antagonistically to determine the afterdepolarization that leads to burst firing. Voltage-gated sodium, calcium and potassium channels all play critical roles in shaping the afterdepolarization that leads to burst firing (Azouz et al. 1996; Metz et al. 2005, 2007). Our data show that FFA effectively decreases repetitive firing in neurons depolarized with long-duration (1 s) current injections (Fig. 6). Importantly, FFA completely abolishes burst firing in physiological conditions as well as in the 4-AP slice epilepsy model (Rutecki et al. 1987; Avoli et al. 1988) in which all CA1 pyramidal neurons are burst firing (Fig. 8). While sodium channel modulation appears to be instrumental in these effects, as suggested by the fact that they are qualitatively reproduced in our simulations in which the I_Na is the only current affected, it is nevertheless possible that actions on other conductances may also have relevant roles in the modulation of neuronal firing by FFA (see below).

Flufenamic acid as potential anti-epileptic tool

Epilepsy results from neuronal hyperexcitability. Pharmacological treatment of epilepsy aims to lessen neuronal excitability either through the potentiation of GABAergic synaptic inhibition (barbiturates and benzodiazepines) or by depressing intrinsic neuronal excitability through modulation of voltage-gated sodium and calcium currents (carbamazepine, valproic acid, phenytoin; summarized in Coulter, 1997). In particular, drugs acting on sodium currents, such as phenytoin and carbamazepine, show a use-dependent effect in consequence of their preferential binding to the inactivated state of the channels (Willow et al. 1985) that results in stabilization of the inactivated state and minimization of the persistent current (Chao & Alzheimer, 1995; Kuo et al. 1997). The increased fraction of inactivated channels limits the number of sodium channels ready to open during repetitive firing, which is believed to occur when the seizure activity spreads (Rogawski & Loscher, 2004). Inhibition of burst firing also represents a mode of action that is shared by several commonly used anti-epileptic drugs (Willow et al. 1985). Thus, FFA modulation of sodium current and cellular excitability appears of potential interest for anti-epileptic therapy. Moreover, the different mode of action suggests that FFA could synergistically complement the existing epilepsy treatments. Additionally, FFA effects may also involve other channel types since FFA has been used for a long time as a blocker of calcium-activated cationic current (Ghamari-Langroudi & Bourque, 2002; Pace et al. 2007) and gap junctions (Srinivas & Spray, 2003). Indeed, previous data show that FFA abolished the sustained afterdepolarization waveform in neocortical slices treated with bicuculline; this effect was attributed to the blockade of calcium-activated cation current (Schiller, 2004), which may act synergistically with the sodium channel modulation to produce anti-epileptic effects. More recently, FFA has been used as an agonist of TREK and TRAAK (Takahira et al. 2005), two channel subunits mediating background potassium currents that are expressed in hippocampal neurons (Fink et al. 1996; Hervieu et al. 2001). Activation by flufenamic acid of these potassium channels would lead to hyperpolarization of the resting membrane potential and represent an additional effect against hyperexcitability. Further studies will be required to investigate the scale and physiological relevance of the FFA effect on the background channels in regulation of hippocampal excitability although the involvement of these channels in the FFA action on CA1 pyramidal neurons in our experimental conditions (recordings obtained at 32°C and with 1 mM EGTA internally) seems minor compared to the Na⁺ current-mediated effect, as suggested by the qualitative reproduction of the FFA effects in a simplified computational model in which FFA acts exclusively on Na⁺ channels. This may be a consequence of the relatively low expression of some other known FFA-sensitive channels in these cells as suggested by the fact that FFA did not induce significant changes in resting membrane potential. This observation is in line with the notion that most of the background conductance in CA1 pyramidal cells is mediated by TASK channels (Taverna et al. 2005). Nevertheless, FFA induced a significant, albeit small, change in input resistance and increased rheobase current about 2-fold. While part of the rheobase increase is reproduced by the simulations, in which the sodium current is the only parameter varied, these data suggest that in pyramidal neurons FFA also affects channels other than sodium. The apparent discordance between the relatively small effect on input resistance and the larger effect on rheobase may suggest that the current involved is not a background current but a current that activates just below threshold; such interpretation is in agreement with the suggested blockade by FFA of a calcium current, as reported by Wang et al. (2006) in the lamprey spinal cord. More studies will be necessary to identify all the currents involved in the FFA-induced rheobase increase in hippocampal pyramidal neurons.

Glossary

Abbreviations

AED

anti-epileptic drug

FFA

flufenamic acid

Author contributions

H.J.-Y.: collection, analysis and interpretation of data; drafting the article. G.B.; collection, analysis and interpretation of data; revising the manuscript. critically for important intellectual content. M.M.; conception and design of the experiments; drafting the article; collection, analysis and interpretation of data. All authors approved the final version of the manuscript.