Valproic acid interactions with the NavMs voltage-gated sodium channel

Significance

Valproic acid is a drug that has been widely used to treat epilepsy and other neurological disorders for many years, but its etiology and site of action are not well known. Among other targets, it has been proposed to bind to and affect voltage-gated sodium channels. In this study, biophysical, electrophysiological, and computational methods have been used to demonstrate that valproic acid binds to the voltage sensor region of sodium channels, at a site distinct from that in the pore domain where hydrophobic anticonvulsant sodium channel-targeting drugs bind, and that it does, indeed, affect channel functioning.

Abstract

Valproic acid (VPA) is an anticonvulsant drug that is also used to treat migraines and bipolar disorder. Its proposed biological targets include human voltage-gated sodium channels, among other membrane proteins. We used the prokaryotic NavMs sodium channel, which has been shown to be a good exemplar for drug binding to human sodium channels, to examine the structural and functional interactions of VPA. Thermal melt synchrotron radiation circular dichroism spectroscopic binding studies of the full-length NavMs channel (which includes both pore and voltage sensor domains), and a pore-only construct, undertaken in the presence and absence of VPA, indicated that the drug binds to and destabilizes the channel, but not the pore-only construct. This is in contrast to other antiepileptic compounds that have previously been shown to bind in the central hydrophobic core of the pore region of the channel, and that tend to increase the thermal stability of both pore-only constructs and full-length channels. Molecular docking studies also indicated that the VPA binding site is associated with the voltage sensor, rather than the hydrophobic cavity of the pore domain. Electrophysiological studies show that VPA influences the block and inactivation rates of the NavMs channel, although with lower efficacy than classical channel-blocking compounds. It thus appears that, while VPA is capable of binding to these voltage-gated sodium channels, it has a very different mode and site of action than other anticonvulsant compounds.

Valproic acid (VPA) (2-n-propylpentanoic acid) is a first-generation antiepileptic drug that has also been used to treat mood, migraine, bipolar, and anxiety among other psychiatric disorders (1, 2). It appears to also exhibit a neuroprotective effect, as observed in some neurodegenerative experimental models (3–6). In humans, the recommended therapeutic total blood VPA concentrations range from 350 to 700 µM for treatment of epilepsy, with the suggested concentration of unbound VPA ranging between 35 and 105 µM in order to avoid significant adverse neurological symptoms in patients with hypoalbuminemia (7). If administrated during pregnancy, VPA has been associated with cognitive deficits, birth defects, and an increased risk of autism, as observed in the clinic (8) and in animal models (9, 10). Despite its use over many decades, there still is no clear information on the mode of action of VPA at the molecular level. Proposed mechanisms have included effects on sodium channel-mediated currents, on GABA receptors, and the efficacy of excitatory neurotransmission (11–14). Early studies on the administration of VPA to neuron cultures indicated its ability to modulate sodium and potassium ion conductance (15) and to modify sodium-dependent action potentials in neurons (16, 17). In neocortical neurons, low concentrations (10 to 30 µM) inhibited persistent tetrodotoxin (TTX)-sensitive sodium currents (18), while inhibition in fast TTX-sensitive sodium channels was observed in hippocampal neurons using high concentrations (500 µM) of VPA (19). Two hundred micromolar VPA was able to inhibit the maximal amplitude of TTX-resistant sodium currents in medial prefrontal cortical neurons, and shift (2 and 200 µM) the channel inactivation curve toward hyperpolarization, causing channel currents to recover more slowly (20). This indicates that while VPA may not completely block voltage-gated sodium channels (VGSCs), it does reduce the fast and transient inward Na⁺ currents, thus interfering with the mechanism of sustained and prolonged firing.

VGSCs are transmembrane proteins, whose openings are associated with the initial stage of propagation of the action potential in excitable cells. Eukaryotic VGSCs are formed from a large α subunit, which alone is responsible for the sodium ion translocation function of the channel, and, in most cases, one or more smaller β regulatory subunits. The α subunit consists of 4 highly similar domains (DI–DIV) with each domain formed of 6 transmembrane-spanning helical segments (designated S1–S6), connected by intracellular and extracellular loops as well as linking regions between domains, plus N- and C-terminal regions. Each domain consists of a voltage sensor subdomain (comprised of the S1–S4 helices) and a pore subdomain consisting of the S5–S6 helices. Prokaryotic sodium channels, in contrast, are composed of 4 identical monomers, each of which corresponds to one of the domains of a human sodium channel. The NavMs prokaryotic sodium channel from Magnetococcus marinus, in particular, has been shown to be both a functionally and structurally relevant exemplar for human sodium channels (SI Appendix, Fig. S1) (21, 22). Indeed, eukaryotic sodium channel antagonists, including antiepileptic and analgesic drugs, bind to and influence the inactivation kinetics of NavMs in parallel manners to their effects on the human sodium channel isoform Nav1.1 (23). Thus, this ortholog has been used as a powerful tool for the study of the nature of the interaction of prospective, as well as current, human drugs, with VGSCs.

It was originally proposed (24) that hydrophobic anesthetics, anticonvulsants, and antiarrhythmic drugs would bind in the inner cavity of the sodium channel pore, blocking the transit of sodium ions between the extracellular and intracellular compartments. Indeed, the location of such a binding site in the central hydrophobic cavity of the pore domain was demonstrated for the NavMs channel (23). That site is adjacent to the channel fenestrations, which provide openings into the pore from the surrounding hydrophobic lipid region (23, 25). However, VPA has very different physical and chemical properties (SI Appendix, Fig. S2) from the highly specific hydrophobic sodium channel-blocking drugs such as lamotrigine, currently used to treat epilepsy, and the local anesthetic lidocaine. The current study has examined the molecular and functional interactions of VPA with NavMs and compared them to known channel-blocking drugs.

Physical methods that have been previously used to determine the effects of ligand binding on sodium channels have included circular dichroism (CD) spectroscopy (to examine whether binding alters the secondary structure of the protein) (26, 27) and thermal melt CD studies to define factors affecting the stability of the protein (28) and the relative stabilities of the transmembrane and intracellular regions of the channels (29). Those studies have generally shown that hydrophobic drug binding increases the stability of both eukaryotic and prokaryotic sodium channels. Crystallographic studies demonstrated that those drugs bind in ways that produce many intermolecular interactions within the large central hydrophobic cavity region of the pore domain (23) and fit within existing pockets in the protein, and thus do not require the protein to refold.

To examine the nature of the interactions of VPA with NavMs sodium channels, we characterized the functional effects (by electrophysiology) and the structural and protein stability effects (by thermal melt synchrotron radiation CD [SRCD] spectroscopy) in both full-length channels and in pore-only constructs. We then identified the location of VPA within the channel by computational docking studies using both the channel and pore structures. These studies indicate on a molecular level that while VPA does interact with this VGSC, both the site and nature of its interaction—in the voltage sensor region, not the central cavity of the pore domain—are very different from the interactions of other anticonvulsant drugs with sodium channels.

Results

The Effects of VPA Binding on NavMs Secondary Structure and Stability (as Indications of Drug Binding).

SRCD spectroscopy has proven to be an effective tool for the study of drug binding to sodium channels and other proteins (30), as it is much more sensitive to small changes than conventional CD spectroscopy, due to the higher penetration of the intense synchrotron light into the samples, which is particularly important for samples in the presence of detergents and/or lipids (31). In this study, the spectra of the full-length and pore-only constructs in the presence and absence of VPA (see Data Availability in the Materials and Methods) were compared (SI Appendix, Methods and Fig. S3). Both constructs exhibited a large positive peak at ∼193 nm and negative peaks at 209 and 223 nm, indicating that at 20 °C the proteins were folded and mostly α-helical (Table 1, “Without VPA,” Top), in accordance with their crystal structures (21, 23). At high temperatures (Table 1, “Without VPA,” Bottom), both structures unfolded to a considerable extent, but as has been noted previously for sodium channels (26, 28, 29) the helical content was not greatly diminished even at the highest temperatures. Upon addition of VPA, the spectra (SI Appendix, Fig. S3) and the resulting calculated secondary structures (Table 1) did not change significantly from those of the apo channel or apo pore-only construct without VPA, at either the lowest or highest temperatures. This is consistent with other observations of drug binding to sodium channels (26, 27) and reflects the robust and stable nature of the structures. Then thermal melt experiments were done to examine whether the presence of the drug influenced the stability of the channel or pore at intermediate temperatures. In the case of the channel (Fig. 1, Top), in the presence (blue) and absence (black) of VPA there were significant differences in the peak magnitudes of the measured CD signals (left panel) at intermediate temperatures—particularly the 193-nm peak, which is most sensitive to changes in helical secondary structure, and in the derived principal-component analyses (right panel), as well as in the calculated secondary structures at the intermediate (62 °C) temperature (Table 1, Left). No such differences in the presence and absence of VPA were detected for the pore (Fig. 1, Bottom; Table 1, Right). These suggest that, in the presence of VPA, the channel, but not the pore structure, is more sensitive to thermal unfolding at intermediate temperatures, even though the structures of the native and fully denatured samples appeared to be the same with and without VPA. Thus, VPA drug binding to the channel does not produce any significant changes in the protein’s structure but causes changes in the stability of the structure. However, the drug does not appear to bind to the pore, as neither a change in structure nor a change in stability occurs in its presence. Interestingly, the direction of change (destabilization) for this hydrophilic drug is in the opposite direction of that often seen for hydrophobic drugs, which have been shown crystallographically to bind in the hydrophobic central cavity of the pore (23) and tend to stabilize prokaryotic sodium channel structures (26, 27); such destabilization effects have been seen for ligands binding to other proteins (32, 33). These results indicate VPA binds in an entirely different location (and hence via a different mode of action) than other sodium channel-active antiepileptic drugs. However, it is notable that another acidic compound has recently been shown (34) to bind to similar regions of the voltage sensor domain (VSD) in a chimeric human/prokaryotic sodium channel.

Table 1.

Calculated secondary structures of the channel and pore in the absence/presence of VPA at 20 °C, 62 °C, and 79 °C

Temperature	Channel		Pore
	% Helix	Δ	% Helix	Δ
Without VPA
20 °C	62 ± 1		69 ± 1
62 °C	41 ± 1	−21	51 ± 1	−18
79 °C	23 ± 0	−39	37 ± 4	−32
With VPA
20 °C	64 ± 0		66 ± 1
62 °C	34 ± 3	−30	50 ± 3	−16
79 °C	21 ± 2	−43	38 ± 1	−28

Values reported are the average of 3 experiments (each of which included 3 measurements at each temperature); ± values represent 1 SD between experiments. “Δ” is a measure of the difference in secondary structure percentages between the given temperature and the initial temperature. The values for the channel with and without VPA at 62 °C (in bold), but not for the pore at this temperature indicate significant differences in structures with and without VPA at the intermediate temperature (but only for the channel, not the pore).

Fig. 1.

Comparison of thermal denaturations of channel and pore constructs in the presence and absence of VPA. (Left) Data showing 193/194-nm peak magnitude changes for the NavMs channel (Top) and the NavMs pore (Bottom) without drug (black squares) and in the presence of VPA (blue circles) as a function of temperature. The curves were normalized to the average of values for 3 independent experiments at 20 °C, and the error bars represent 1 SD in the measurements. (Right) Principal-component analyses of the thermal denaturation data. Fractions of the first basis spectra that contribute to the net SRCD spectrum with increasing temperatures from 20 to 79 °C for the NavMs channel (Top) and the NavMs pore (Bottom) without drug (black squares) and in the presence of VPA (blue circles). The curves were normalized so that the highest value for the first component is 1.0. The error bars represent 1 SD in the measurements of independent experiments. VPA clearly influences the stability of the channel in the intermediate temperature range, while it does not influence the stability of the pore in the same range. This indicates that VPA binds either to the VSD (or possibly to the interface between the VSD and pore domain), but not within the pore domain.

Electrophysiology Characterization of VPA on NavMs in HEK293t Cells.

To examine whether the effect seen of VPA on NavMs thermal stability is reflective of ion channel antagonism, whole-cell patch-clamp experiments were conducted on HEK293t cells transiently transfected with plasmids encoding for the channel, and the impact of VPA on sodium current was measured (Fig. 2). Under tonic inhibition, VPA dose-dependently reduced the NavMs sodium current (Fig. 2A). The half inhibition concentration (IC₅₀) of VPA for the sodium current was 691 μM (Fig. 2B), which approximates its reported potency against neuronal sodium channels (15) and is within the targeted free plasma concentration (350 to 700 µM) for use as an anticonvulsant (7).

Fig. 2.

VPA blocks NavMs sodium currents by enhancing the inactivated state. (A) The onset of NavMs sodium current (I_Na) by the indicated concentrations of VPA. (Inset) Corresponding exemplar whole-cell current responses before and after drug block. HEK293t cells expressing NavMs were held at −180 mV, and sodium currents were activated by a 0.2-Hz train of 50-ms depolarizations to −20 mV. (B) VPA concentration–current block relationships (error ± SD; N = 4 cells). The inhibition concentration at 50% block (IC₅₀) was estimated by fitting the average percent current block to the Hill equation. (C, Top) The voltage protocol used to activate sodium currents before and after exposure to 1 mM VPA, and (C, Bottom) results of this protocol. (D) The effect of VPA on NavMs voltage-dependent activation and inactivation. The resulting voltage-dependent activation (circles) and inactivation (squares) before and after addition of 1 mM VPA (error ± SEM; n = 4 cells). (E) Open probability (Po) effects: the impact of VPA on the NavMs window current (available channels over membrane potential).

To examine VPA’s mechanism of action, the impact on steady-state voltage-dependent properties of NavMs sodium currents (Fig. 2C) was tested. At VPA concentrations <1 mM, the voltage dependence of activation was not significantly different from that of the control. However, a 12-mV negative shift in the voltage dependence of inactivation was observed (Fig. 2D), consistent with effects observed in neural sodium currents measured from rodent neurons (19, 35). By integrating the difference of the activation and inactivation Boltzmann relationships (Fig. 2E), it appears that VPA reduces the open probability (Po) at potentials from −129 to −27 mV, which spans the range of resting membrane potentials found in excitable cells such as neurons and myocytes. These results suggest VPA enhances the inactivated state of NavMs, ultimately reducing the number of available channels that can conduct sodium currents.

Computation Docking of VPA to Channel and Pore Structures.

Computational docking is a useful tool for identifying potential binding sites of ligands and drugs in protein structures. Thus, they have been extensively used for the rational design of therapeutic compounds and identification of new chemotypes. In this study, docking studies were undertaken to identify the binding sites for VPA. Because of the SRCD studies, it was anticipated that binding sites in the channel structure would involve the voltage sensors rather than the pore domains. Nevertheless, to test this, parallel studies were undertaken using both the NavMs channel (21) and pore (36) crystal structures.

Initially “blind” docking, using Autodock software (37), was used to identify possible binding sites on the surface of the proteins, followed by a second docking round to assess the best location of VPA in the principal identified binding sites. For the NavMs channel structure, the 2 tightest binding sites identified were in the VSD: 1) at the extracellular end of the VSD, composed of S1–S4 helices, between helices S3 and S4 (−6.2 kcal/mol) (Fig. 3A, site 1), and 2) at the intracellular end of the VSD (−4.5 kcal/mol) (Fig. 3A, site 2) between helices S1, S2, and S4. For the pore construct that does not have a VSD, the top site (Fig. 3B) was located at the extracellular surface above the entrance to the selectivity filter (−3.3 kcal/mol), not nearly as tight as either of the binding sites in the channel structure. This is consistent with the experimental SRCD observations that the channel VSD contains the primary binding site(s) for VPA. No significant sites were found for either construct in the hydrophobic internal cavity of the transmembrane ion pathway below the selectivity filter, the region where hydrophobic channel blocker compounds have been shown to bind (23). The docking analysis was repeated with alternative docking and binding site identification methods (SI Appendix, Methods), which produced very similar results (SI Appendix, Fig. S4). Furthermore, molecular-dynamic simulation approaches (SI Appendix, Methods and Fig. S9) have indicated that the computationally identified VPA binding site and pose is maintained during the trajectories, further reinforcing the validity of the prediction.

Fig. 3.

Docking of VPA in the NavMs channel and pore structures using AutoDock, with detailed views of the VPA binding sites. (A) Channel: The top VPA binding sites identified in the channel structure (calculated energies of sites 1 and 2, respectively, are −6.2 and −4.6 kcal/mol) are only located in the VSD. (B) Pore: The top VPA binding site (calculated energy of −3.3 kcal/mol) identified in the pore structure (Left) is in the selectivity filter and between pore helices S5 and S6 in the pore domain. The 4 monomers of the tetramer are depicted in different colors using Pymol software. For comparison, docking results using an alternative procedure, GlideSP, are shown in SI Appendix, Fig. S4. Both types of docking essentially produced the same results for the binding sites. Hence in all other figures in SI Appendix, only the Autodock results are shown.

For comparison, the recently published cryoelectron microscopy structure of the human Nav1.2 sodium channel (38) was also subjected to similar docking simulations with VPA (SI Appendix, Table S1 and Figs. S5 and S6). This structure was chosen as the target exemplar from among the recently determined human Nav structures because it is the only one of the human sodium channels solved thus far from a brain-localized channel (i.e., a target protein for antiepileptic drugs). Sequence identities and superpositions of NavMs and the human Nav1.2 structures (SI Appendix, Fig. S1) indicate that they are highly similar; the tightest VPA binding appears to be to domain 3 in Nav1.2 (SI Appendix, Fig. S6C). Indeed, the top docked binding sites in the NavMs and Nav1.2 structures were found to be very similar (SI Appendix, Figs. S5 and S6). These results also indicate that the hydrophilic VPA has a tendency to bind in the VSDs of sodium channels, not in the hydrophobic cavity within the pore domain where hydrophobic analgesic and antiepileptic compounds bind (23, 25).

The main docking site for VPA in NavMs was also compared to the docked site (SI Appendix, Fig. S7 and Table S1) for a hydrophilic drug compound (5P2) in the chimeric Nav structure, which consists of the extracellular half of one VSD of the human Nav1.7 channel (which includes the drug binding site) replacing the equivalent region in the structure of the prokaryotic sodium channel NavAb (Protein Data Bank [PDB] ID 5EK0) (34), a close homolog of NavMs. The site identified for 5P2 docking was also in the VSD of the chimera at a very comparable location (SI Appendix, Figs. S7 B and C and S8) to both the NavMs and Nav1.2 docked VPA sites. Then as a control for the docking procedure, the docked site was compared with the experimentally identified site (34) for the drug in the crystal structure (SI Appendix, Fig. S7 B and C and Table S1). That the docked VPA and 5P2 crystal structures were very similar gives credence to the docking procedures and is further suggestive that hydrophilic compounds such as VPA bind to sodium channels, but in different manners than do other classes of antiepileptic drugs.

Conclusions

VPA is a branched short-chain fatty acid, which is converted into its active form, a valproate ion, in the blood, and has very different physical and chemical properties from the highly specific hydrophobic sodium channel-blocking drugs such as lamotrigine, used in the treatment of epilepsy, and local anesthetics such as lidocaine. Those drugs have been shown to bind to, and block ion passage through, the hydrophobic central channel of the pore domain that connects the cell exterior and interior (23, 25).

The first evidence for the anticonvulsant activity of VPA was suggested more than 3 decades ago, but the nature of its interactions with sodium channels have remained unknown. The present study has illustrated VPA binding to sodium channels and its ability to interfere with the inactivation process at concentrations near to therapeutic values. The relatively low binding affinity of VPA for sodium channels may be relevant for future therapeutic considerations: In a clinical setting, VPA administration tends to be at high concentrations, which can elicit significant side effects, such as hepatotoxicity, mitochondrial toxicity, neurological toxicity, adverse metabolic and endocrine events, impairments in normal development during pregnancy related to autism spectrum disorders, and teratogenicity among others (35). These could arise from its nonspecific (or less specific) binding to a wide range of channels in different tissues.

In this study, thermal stability SRCD studies used to discern whether VPA interacts with either the pore region or elsewhere in the NavMs channel, showed that while the net secondary structure conformations of the NavMs channel and pore are not changed in the presence of VPA, the thermal stability profile of the channel, but not the pore-only construct, is influenced by the presence of the drug. Its influence is to destabilize the channel, the opposite effect of that observed for other sodium channel-blocking drugs, which increase the stability of the sodium channel pore domain (26, 27). Comparisons of the thermal effects on the full-length channel with the lack of effect on the pore-only construct suggested that the site of VPA interaction was either in the VSD or in the interfacial region between the 2 domains; neither of these sites has been shown to be the binding site for hydrophobic channel-blocking drugs. Note that it was not possible to do the converse experiment (comparing the effects on the VSD alone with those of the channel) because the VSD on its own does not form a stable tetrameric structure. Nevertheless, the channel versus pore differences seen in this study are sufficient to indicate the primary site of VPA binding is not in the same region as other hydrophobic channel-blocking drugs. Molecular-docking studies indicated where in the channel the drug might bind, and whether that site as compatible with the thermal stability studies. The primary sites identified for VPA docking in both prokaryotic and eukaryotic sodium channel structures were all in the voltage sensor region between helices, which could produce partial decoupling of the closely associated transmembrane regions, suggesting a new site for targeted drug development.

In summary, the combination of experimental and computational studies in this work has indicated that the primary target sites for VPA binding in sodium channels are in the VSD, not the pore domain, far removed from the central hydrophobic transmembrane cavity that has been identified as the primary binding site for other sodium channel antiepileptic and analgesic drugs.

Materials and Methods

The full-length NavMs sodium channel (21) and the pore-only construct (23) were expressed and purified as previously described. SRCD spectra were collected at the ISA synchrotron (Denmark), with replicate measurements later obtained at the Soleil Synchrotron (France), and the KARA synchrotron (Germany). Principal-component analyses were carried out using CDToolx software (39) and secondary structure analyses used the Dichroweb server (40).

Whole-cell patch-clamp measurements on NavMs channels expressed in HEK293t cells were performed as previously described (23).

Docking calculations used the crystal structures of the NavMs channel (PDB ID code 5HVD) and NavMs pore (PDB ID code 4P9O), the cryo-EM structure of the human Nav1.2 channel (PDB ID code 6J8E), and the crystal structure of a prokaryotic NavAb and a human Nav1.7 chimeric channel (PDB ID code 5EK0).

Data Availability.

SRCD spectra and metadata have been deposited in the Protein Circular Dichroism Data Bank (PCDDB) (41, 42) located at http://pcddb.cryst.bbk.ac.uk under the codes CD0006224000–CD0006224008, CD0006225000–CD0006225008, CD0006226000–CD0006226008, and CD0006227000–CD0006227008 (43). The CD analysis software package (CDToolx) is freely available as a downloadable package at http://www.cdtools.cryst.bbk.ac.uk.

The plasmid for NavMs (originally described in ref. 21) is available upon request from P.G.D. (paul.decaen@northwestern.edu).