Cannabidiol Interactions with Voltage-Gated Sodium Channels
1Institute of Structural and Molecular Biology, Birkbeck College, University of London, London WC1E 7HX, U.K
2GW Research Ltd., Cambridge, U.K
+Author to whom correspondence should be addressed; email: b.wallace@mail.cryst.bbk.ac.uk.Abstract
Voltage-gated sodium channels are targets for a range of pharmaceutical drugs developed for treatment of neurological diseases. Cannabidiol (CBD), the non-psychoactive compound isolated from cannabis plants, was recently approved for treatment of two types of epilepsy associated with sodium channel mutations. This study used high resolution X-ray crystallography to demonstrate the detailed nature of the interactions between CBD and the NavMs voltage-gated sodium channel, showing CBD binds at a novel site at the interface of the fenestrations and the central hydrophobic cavity of the channel. Binding at this site blocks the transmembrane-spanning sodium ion translocation pathway, providing a molecular mechanism for channel inhibition. Modelling studies illuminate why the closely-related psychoactive compound THC may not bind to these channels. Finally, comparisons are made with the TRPV2 channel, also recently proposed as a target site for CBD. In summary, this study provides novel insight into a possible mechanism for CBD with sodium channels.
Introduction
Voltage gated sodium channels (Navs) specifically enable the passage of sodium ions across cell membranes, contributing to the electrical signalling in cells (Ahern et al, 2016). The nine homologous mammalian sodium channel subtypes, designated hNav1.1-hNav1.9 (Supplementary Figs. 1 & 2), have different functional characteristics and expression profiles within different tissues (Catterall et al, 2005). Mutations of hNavs have been associated with a range of channelopathies, including pain, epilepsy, and heart disorders, making them major targets for drug development (Bagal et al, 2015, Kaplan et al, 2016).
Cannabinoids (including cannabidiol (CBD) and tetrahydrocannabinol (THC)) are hydrophobic compounds (Supplementary Fig. 3) produced by the cannabis plant. Whilst THC has been primarily associated with psychoactive drug use (Rosenberg et al, 2015; Pisanti et al, 2017), the non-psychoactive component, CBD, has been extensively investigated for its potential clinical applications as a therapeutic drug for treatment of epileptic conditions (Rosenberg et al, 2017). One formulation of CBD (GW Research Ltd., UK) has recently been approved by the European Medicines Agency and the Federal Drug Administration for use in children for the treatment-resistant epilepsies Dravet Syndrome and Lenox-Gastaut Syndrome (Sarker & Nahar, 2020). Both of these diseases are rare early onset epilepsies associated with Navs, with Dravet patients often having mutations in the hNav1.1 human sodium channel gene SCN1A (Marini et al, 2011). Furthermore, CBD has been shown to attenuate seizures and social deficits in a mouse model of Dravet syndrome (Kaplan et al, 2017). Despite a significant amount of evidence reporting on the effectiveness of CBD for treating epileptic conditions (Cross et al, 2017; Devinsky et al, 2018), the molecular basis of its target interactions remain unclear (Watkins, 2019).
Electrophysiology studies, however, have indicated that CBD can modify sodium channel functioning: at 10 μM, it significantly reduced action potentials in rat CA1 hippocampal neurons, as well as the Nav current density in human blastoma cells and mouse cortical neurons (Hill et al, 2014). In addition, CBD has been shown to inhibit the channel activities of human Nav1.1 to Nav1.7 isoforms, as well as those of the prokaryotic Nav homologue NachBac, with IC50s ranging from 1.5 to 3.8 μM, which suggests inhibition at physiologically-relevant concentrations. Functional studies (Patel et al, 2016; Ghovanloo et al, 2018) on both hNavs and the homologous prokaryotic Navs suggested CBD interferes with the inactivation processes of these channels. In a recent study (Mason & Cummins, 2020) CBD was shown to inhibit both resurgent and persistent sodium currents of hNav1.2 at concentrations of 1 μM. However, to date, the structural basis of the interactions of CBD and sodium channels have not been identified on a molecular level.
CBD has also been suggested to be a potential inhibitor of the Transient Receptor Potential Cation Channel Subfamily V Member 2 (TRPV2) channel (Qin et al, 2008; Morelli et al, 2013), which facilitates the non-specific movement of both sodium and calcium ions through plasma membranes. According to electrophysiology studies, CBD activates rat TRPV2 with an EC50 of 3.7μM (Qin et al, 2008), although the link with epilepsy (Morelli et al, 2013) is much less direct than that for sodium channels. Recently cryo-electron microscopy (cryo-EM) was used to elucidate the structure of TRPV2 in a CBD-bound state at a nominal resolution of 3.2 Å (Pumroy et al, 2019); that study indicated the presence of CBD in the pore region of the protein structure, thus supporting the proposal for TRPV2 being a candidate target for CBD binding.
In the present study, in order to examine the nature of the interactions of CBD with sodium channels, the high-resolution crystal structure of a complex of CBD with the NavMs voltage-gated sodium channel from M. marinus (Sula et al, 2017), was determined, enabling the binding sites for the CBD molecule to be clearly defined at high resolution. The NavMs channel has been shown to be a good exemplar for hNavs as they exhibit not only functional (Bagneris et al, 2014; Ulmschneider et al, 2013; Ke et al, 2018), but also sequence and structural homologies (Supplementary Figs. 1 and 2) (Sula et al, 2017; Sula & Wallace, 2017). The NavMs-CBD structure described in this work hasenabled comparisons of that binding site with binding sites of other sodium channel ligands and modifiers in human sodium channels, including the hNav1.1 and hNav1.2 channel, which are the predominant hNav isoforms found in human brain tissue. It also provides a means for comparing its binding site for CBD with that found in the TRPV2 channel.
This high resolution crystallographic study of a sodium channel–CBD complex thus provides a means of both understanding the molecular interactions of CBD and sodium channel targets, and how these may be related to its use for treatment of epilepsy.
Materials and Methods
Materials
Thrombin was purchased from Novagen Inc (Germany), decanoyl-N-hydroxyethylglucamide (Hega10) was purchased from Anatrace (USA), and dimethyl sulfoxide (DMSO), sodium chloride, 2-amino-2-(hydroxymethyl)-1,3-propanediol (Tris), and imidazole were purchased from ThermoFisher Scientific (USA). Purification columns were purchased from GE Healthcare (USA). Cannabidiol (CBD) samples were supplied by GW Research Ltd. (UK). The F208L (NavMsL) mutation was introduced using the SLIM site-directed mutagenesis protocol (Chiu et al, 2004), using the forward primer 5′-CTCACCACCCTGACCGTGCTCAACCTGTTTATTGG-3′ and reverse primer 5′-GAGCACGGTCAGGGTGGTGAGCATGATGAACGGGATG-3′. The sequence was verified by Source Bioscience, UK.
Protein expression and purification
The NavMs (Uniprot ID A0L5S6) and NavMsL proteins were expressed and purified as previously described (Sula et al, 2017), with the following modifications: the bound protein was eluted in a buffer containing 20 mM Tris, pH 7.5, 300 mM NaCl, 0.5 M imidazole and 0.52% Hega10. The Histag was removed by thrombin cleavage overnight at 4° C. The protein sample was loaded onto a Superdex 200 column and eluted with 20 mM Tris, pH 7.5, 300 mM NaCl, and 0.52% Hega10 buffer. Protein samples were pooled and concentrated to 10 mg/ml using a 100 kDa cut-off Amicon concentrator and stored at a concentration of 10 mg/ml at −80 °C.
Crystallisation, Data Collection and Structure Determination
1 μl of cannabidiol (100 mM) in 100% DMSO was added to 50 μl of the purified protein solution to produce a final protein concentration of ~10 mg/ml containing 2 mM CBD and 2.5% v/v DMSO. The best crystals were grown at 4 °C via the sitting drop vapour diffusion method using a 2:1 ratio of the protein and reservoir solutions containing 0.1 M lithium sulphate, 0.1 M HEPES, pH 7, and 40% v/v PEG200. The apo NavMsL crystals were grown under the same condition as the crystals of the CBD complex, but without the DMSO and drug. Crystals were flash-frozen, with the PEG200 acting as the cryo-protectant.
Data were collected on beamline P13 at the Electron Synchrotron (DESY, Germany); on beamline Proxima1 at the Soleil Synchrotron (France), and on beamlines IO3, IO4, and I24 at the Diamond Light Source (UK). Hundreds of crystals were screened and full data sets were collected from more than 40 crystals. Diffraction images were integrated and scaled using XDS (Kabsch, 2010) and then merged with Aimless (Evans & Murshudov, 2013) using the CCP4 suite of programmes (Winn et al, 2011). The structure was determined from the crystals which diffracted to the highest resolution (2.2 Å for the apo protein, and 2.25 Å for the CBD complex). Because of the small but significant variations in the unit cell dimensions and resolution between different crystals of the same type produced under the same conditions, as we have seen previously (Naylor et al, 2016; Sula et al, 2017), datasets from different crystals were not merged.
The structure determinations by molecular replacement were as previously described (Sula et al, 2017) using Phaser (McCoy et al, 2007) with the full-length wildtype NavMs structure (PDB 5HVX) as the search model. Model building was carried out using Coot (Emsley et al, 2010). Refinement was done using REFMACS (Murshudov et al, 2011). Data collection, processing and refinement statistics for both the apo and CBD complex structures are listed in Supplementary Table 1. The structure quality was checked using PROCHECK (Laskowski et al, 1993) and MolProbity (Chen et al, 2010), which indicated that 99.2% of the residues were in allowed conformations. Figures were created in CCP4mg (McNicholas et al, 2011), unless otherwise noted.
Results
This study utilised the prokaryotic NavMs voltage-gated sodium channel, which has been previously shown to be an excellent structural and functional exemplar for human sodium channels, to examine the site of interactions of the naturally-occurring non-psychoactive CBD compound isolated from cannabis plants. One formulation of CBD (GW Research Ltd., UK) has recently been approved by the European Medicines Agency (EMA) and the Food and Drug Administration (FDA) (Sarker & Nahar, 2020) for treatment of specific and severe epilepsies. NavMs was not only chosen for this study because it exhibits both strong sequence and structural homology to hNavs (Supplementary Figs. 1 & 2), but it has also been shown to have highly similar functional (Bagneris et al, 2014), conductance (Ulmschneider et al, 2013), and drug binding (similar IC50 values) characteristics (Bagneris et al, 2014) as human Nav1.1 sodium channels. Whilst NavMs channels are tetramers with each monomer consisting of 6 transmembrane (TM) helices (4 of which form each of the voltage sensor subdomains and 2 of which form the pore subdomains), all of the hNav channel isoforms are monomers of four similar but not identical domains (each of which consists of 6 transmembrane helices that are comprised of 4-helical voltage sensor subdomains and 2-helical pore subdomains). The major difference between NavMs and the hNavs is the presence of the inter-domain loop regions in the human channels, (which also differ considerably between hNavs) (Supplementary Fig. 1, Fig. 5).
The major reason for using crystal structures of the NavMs channel for this study is that they provide the, to date, highest resolution (~2.2-2.5 Å) views of any sodium channel (Naylor et al, 2016; Sula et al, 2017), especially of the TM and drug binding regions, thus enabling detailed views of the protein molecular structures with drugs bound to them. In contrast, the cryo-EM structures of hNavs available to date generally have overall resolutions of between ~4-5 Å, with the transmembrane regions having the best resolutions of ~3 Å, with their extra- and intra-membranous regions being less well defined. However, these similarities would not be sufficient to indicate the value of using NavMs for understanding the molecular basis of drug interactions if their functional roles (conductance and drug binding affinities) were not comparable to those of hNavs. NavMs and hNav1.1 exhibit similar ion flux and conductance properties, and as well as very similar binding affinities for a wide range of sodium channel-specific drugs (Bagneris et al, 2014).
Structure/function/drug-binding studies using some of the other prokaryotic sodium channels have also showed their comparability to hNavs for drug binding (Payandeh et al, 2011; Jiang et al, 2019), and they too have been used for drug discovery projects (Martin & Corry, 2014; Ouyang et al, 2007), although their structures tend to be of lower resolution than those of NavMs. Much of the focus of this study has been on comparisons of NavMs with hNav1.2 and hNav1.1, as these are the sodium channels primarily found in human central nervous system tissues. Although as yet there is no structure available for the hNav1.1 channel, its strong sequence homology to the hNav1.2 has enabled the production of the molecular model used in the comparison studies described her Fig. 5).
The CBD Binding Site in Sodium Channels
The CBD binding site is located and clearly visible in a well-defined region of the NavMs-CBD structure (Figs. 1A & B). It is sited in a hydrophobic pocket present in each subunit that runs perpendicular to the channel direction (Montini et al, 2018) (such features have been designated “fenestrations” and are located (horizontally in Fig. 1C) in the TM region, just below the level of the selectivity filter, and are the features originally proposed by Hille (1977) as sites for ingress of hydrophobic drugs into the channel interior. CBD is located at the end of the fenestration that lies closest to the central pore, and protrudes into (and blocks) the central transmembrane cavity, just below the sodium ion selectivity filter (Fig. 1C). There is enough room for four CBD molecules in this region, although one would be sufficient to block sodium ion passage, as seen from the HOLE (Smart et al, 1993) depictions (Fig. 2A and Supplementary Fig. 4) and pore radius plots (Fig. 2B), which show the size of the transmembrane pathway with and without different numbers of CBD; the blockage clearly provides a mechanism for channel inhibition as well as a basis for understanding the concentration-dependence of the drug effects. Each binding site is comprised of 11 residues from three different subunits in NavMs (shown in different colours in Fig. 3). The corresponding residues in hNavs are from three different domains of the same polypeptide chain (Figs. 4 & 5, Supplementary Fig. 5).
The location of this binding site is very close to the locations of the binding sites that have been identified for analgesic and other hydrophobic compounds in both NavMs (Bagneris et al, 2014) (Supplementary Fig. 6) as well as in another bacterial sodium channel, NavAb (El-Din et al, 2018). This is of interest because those other compounds also inhibit hNav functions, and so suggest the importance of this site for drug interactions in humans. All of the interactions seen except one, that of residue M175 (Fig. 3B), involve hydrophobic interactions rather than hydrogen-bond formation (but that particular interaction between the main chain carbonyl of residues M175 and the OH group present in CBD, may be important for specificity of binding – see next section). It should be noted here, that this region of the apo structure also exhibits some electron density (but has a different size and shape) that has been attributed to detergent molecules, The (2Fo-Fc) electron density map of the CBD complex (Fig. 3A) clearly indicates that in these crystals the site is occupied by CBD rather than detergent.
The Molecular Basis of CBD Binding Relative to THC
There are two main cannabinoids that can be extracted from cannabis plants, the psychoactive tetrahydrocannabinol (THC) and the non-psychoactive cannabinoid (CBD). Electrophysiological studies on hNavs and the bacterial NachBac have identified CBD (Ghovanloo et al, 2018, Patel et al, 2016) as having functional effects that are distinct from those of THC on these channels.
The chemical structures of CBD and THC are very similar (Supplementary Fig. 3), differing only by the presence of an additional free hydroxyl group on one of the rings in CBD (in THC the equivalent oxygen forms part of a closed pyran ring). Therefore, the structure of the CBD/NavMs complex was examined to see if it could provide a clue as to the reasons for the different functional effects of the two compounds. As can be seen in Supplementary Fig. 7, by placing the THC structure into the CBD binding site with the same orientation as found for CBD, it can be physically and sterically accommodated. However, and crucially, it is missing the one electrostatic interaction seen between CBD and NavMs: the hydrogen bond between the oxygen of the main chain residue M175 and the drug. This is the consequence of the absence of the additional free hydroxyl group in THC, as noted above. That hydroxyl group is the one which forms the hydrogen bond present in the CBD-protein complex. This provides an additional intermolecular interaction for CBD, and could account for the differences in binding affinities of the two compounds (Ghovanloo et al, 2018) [as well as (possibly) the differences in psychoactive properties of the compounds].
Specificity/Potential Interactions with Other hNavs
The focus of functional effects of CBD on hNavs has primarily been on hNav1.1, due to its association with epilepsy, although there is yet no structure for this isoform. However, it has been possible to examine potential interactions using a hNav1.1 homology model based on the hNav1.2 cryo-EM structure (Supplementary Fig. 5), which suggests, not surprisingly, that the interactions would be very similar to those of Nav1.2. In NavMs, the involvement of T207 residue is of importance as it is well established as the primary binding site for local anaesthetics. Furthermore, when the equivalent residue (F1774) was mutated in hNav1.1 the binding affinity of CBD was found to decrease by a factor of 2 (Ghovanloo et al, 2018). The binding site residues (coloured red in Fig. 4) include both residues that are identical/homologous in NavMs and hNavs as well as residues that are only found in NavMs and not in human Navs. In most cases the non-cognate residues are also variable between hNavs and would thus appear not to be essential for the interactions.
Comparison with Binding to the TrpV2 Channel
A recent study (Pumroy et al, 2019) has described the interaction of CBD with the TRPV2 channel, as demonstrated by a cryo-EM investigation of its complex, solved at an overall resolution of 3.2 Å. The location of the CBD was visible in the structure, as were the general features of the binding site. Although the lower resolution of that structure did not allow detailed analysis of its binding site, it was clear that involved a number of hydrophobic side chains, and required a partial refolding of the adjacent region of the protein polypeptide. The binding site found for CBD in the TRPV2 structure is in a similar region to that of CBD in NavMs (Fig. 6). However, the sodium channel CBD site is located further into the fenestration than it is in TRPV2, but closer to the ion binding sites and thus would more effectively block the transmembrane passageway for ion conductance.
Conclusions
This study has demonstrated the nature of the interactions of CBD and a voltage-gated sodium channel, showing that CBD binding blocks the transmembrane pathway for sodium ion translocation through the membrane (Naylor et al, 2016), and hence provides a potential mechanism for the functioning of CBD in sodium channels. This further suggests a possible molecular basis for the medicinal effects of CBD in the treatment of epilepsies, as sodium channels have been shown to be causally-related to various types of human epilepsy, with disease-related mutations interfering with sodium ion transmembrane flux. The CBD binding site is a novel site, near to, but not coincident with, known analgesic binding sites in sodium channels; binding at this site would effectively block sodium channel functioning. The binding site is located at the pore end of the transmembrane fenestrations which enable the ingress of hydrophobic molecules into the channel lumen, hence indicating this may also provide the pathway for CBD to enter and block the channels.
Examination of the residues involved in the binding site interactions and modelling of the THC into the CBD binding site have indicated a possible reason for why the closely-related psychoactive cannabinoid THC, has not been observed to have a similar effect on epilepsy nor on sodium channel function: THC would be able to physically fit in the site when oriented in the same manner, but it does not have the same hydroxyl moiety that in CBD forms an important hydrogen-bonding interaction with the channel protein.
Recent cryo-EM structural studies (at lower resolution) have suggested that the TRPV2 channel may be the CBD binding target, although that study did not show the relationship of the binding site to epilepsy-based mutations. However, whilst the TRPV2 channel has a quite different overall fold from that of sodium channels and it acts as a conduit for much larger substrates, it is interesting that the binding site for CBD in TRPV2 appears to be in a roughly comparable structural feature near the transmembrane substrate pathway to that found in this study for the ion pathway in sodium channels.
In summary, this study has provided high resolution structural evidence for the basis of the molecular interactions of CBD, a drug recently approved for treatment of epilepsy, with a voltage-gated sodium channels target. The described structural work can therefore guide further functional studies to explore differential CBD selectivity for human Nav isotypes and their relevance to clinical studies, thus shedding further light on the polypharmacological profile of CBD.
Competing Interests
BJW and RRR are employees of GW Research Ltd and own share options in GW Pharmaceuticals plc. All other authors declare no competing interests.
Funding acquisition
Grants BB/L006790 and BB/R001294 from the U.K. Biotechnology and Biological Science Research Council [BBSRC] (to BAW). Ph.D. studentship from the UCL-Birkbeck Medical Research Council DTP programme (to LGS). Beamtime grants for access to the DESY (Germany), Soleil (France) and Diamond (UK) synchrotrons (Birkbeck/UCL BAG consortium). The funding sources were not involved in the study design, data collection and interpretation, or decision to submit the work for publication.
Datasets/Availability
Structure factors and coordinates for the apo and CBD-bound NavMsL structures have been deposited in the Protein Data Bank under accession codes PDB6YZ62, and PDB6YZ0, respectively.