Voltage-Gated Sodium Channels in Drug Discovery

1.1. Membrane potential, Nernst and Goldman equations

In a typical cell, sodium, potassium, chloride, and other membrane permeable ions are in unequal distribution across plasma membrane, Figure 1A. This unequal distribution and its resultant electrical gradient can be explained by Donnan equilibrium. For a specific ion, the electrical potential difference that exactly counterbalances diffusion due to the concentration difference is called the equilibrium potential for that specific ion. To use Na+ as an example, at equilibrium, the chemical force moving Na+ into the cell is balanced by the electrical force moving Na+ out of the cell (Figure 1B).

For each ion, the equilibrium (or reversal) potential, Eion, can be calculated by the Nernst equation (Eq. (1)), where R = gas constant, 8.135 J K⁻¹ mol⁻¹; T = temperature in K (273 + temp in ^oC); z = valency of ion (Na⁺ = 1, Ca²⁺ = 2, and Cl⁻ = −1); F = Faraday’s constant, 9.684 × 104 C mol⁻¹. In a typical mammalian neuron with [Na⁺], [K⁺], and [Cl⁻] described in Figure 1A, based on Nernst equation, we can calculate ENa= 67 mV, EK= −83 mV, and ECl= −61 mV. In cell resting state, the experimental measured membrane potential Em= −65 mV. As indicated in Figure 2, Na⁺ has a tendency to flow into the cell due to Em<ENa, while K⁺ flow out of the cell (Em>EK), and Cl⁻ near equilibrium (Em≈ECl). These concentration and electrical gradients are maintained by the dynamic equilibrium of ion channels and active ion transporters, most importantly by sodium pump (Na/K-ATPase).

The whole cell membrane potential Emcan be calculated by all permeable ions’ equilibrium potentials Eionand their relative permeability P, which is described by Goldman-Hodgkin-Katz equation (GHK equation) (Eq. (2)). This equation explained the experimental finding that resting membrane potential is more depending on P_K, which is about 25-folds to P_Na.

1.2. Hodgkin-Huxley model

The Hodgkin-Huxley model describes how action potentials in neurons are initiated and propagated [4]. Originally developed to fit action potential dynamics of squid giant axon, this model has been successfully applied to a wide range of neurons. It describes the electrical properties of excitable membranes as typical electrical circuit components. For instance, the cell membrane is modeled as a capacitor with capacitance (C_m) and ion channels are resistors with conductance of Na channel (gNa), K channel (gK),and leak channel (g¯L).

From Hodgkin-Huxley model, the total cell membrane current, I, can be calculated by Eq. (4), where membrane potential Vm, ion conductances gNaand gK, sodium activation variable m, sodium inactivation variable h, potassium activation variable n are variable functions of time, whereas ENa, EK,EL, CM,and g¯Lare constants.

Hodgkin-Huxley model provides a relatively simple and experimentally testable equation to deduce Na⁺ conductance change with time and voltages (Figure 2B), and nerve action potential (Figure 2C). It also embodies the three key features of Nav channels: voltage-dependent activation (submillisecond scale), rapid inactivation (millisecond scale), and selective Na⁺ conductance.

1.3. History and basic concepts of electrophysiology

Electrophysiology was originated by Luigi Galvani in studying “animal electricity” in 1780s. In 1903, an extracellular recording technique, electrocardiography, was invented by Willem Einthoven (1924 Nobel Laureate); in 1952, intracellular recording technique was developed by Alan Hodgkin and Andrew Huxley (1963 Nobel Laureates); and in 1976, Erwin Neher and Bert Sackmann (1991 Noble Laureates) succeeded in measuring the ionic current of single channels in the cell membrane by developing patch clamp technique. These works were fundamental in revealing the physiological function of ion channels. Since then, the field of electrophysiology had undergone rapid evolution, especially after the introduction of automated patch clamp in the early 2000s.

Manual patch clamp technique employs glass microelectrode(s) with desired filling solution and tip diameter to perform either voltage or current clamp. For larger cells, such as Xenopus oocytes, two-electrode voltage clamp (TEVC) is performed using two electrodes: one to measure membrane potential and the other to apply the current, with each tip diameter <1 μm resulting in 10–100 MΩ resistances. For most other circumstances, a single electrode with an open tip diameter 1–3 μm (1–3 MΩ resistances) is used for whole cells or small patches of cell membrane recording. A typical patch-clamp recording starts with cell attaching procedures including: placing glass tip next to a cell, using gentle suction, drawing a piece of the cell membrane to the microelectrode tip, and then letting glass tip to form a high-resistance seal with the cell membrane (ideally >1GΩ, so-called “gigaseal”). By applying different following-up manipulations, the patch-clamp can be achieved in four configurations: cell-attached patch, whole-cell patch, inside-out patch, and outside-out patch, as shown in Figure 3.

Many other electrophysiological techniques have been developed for various applications, including extracellular recordings such as electroencephalography (EEG), electrocardiography (ECG or EKG), and electromyography (EMG) for clinical diagnosis; artificial lipid bilayer recording for studying activities of reconstituted ion channel proteins; automated patch clamp recording to enable high-throughput recordings; and optogenetics to employ light to switch on and off ion channel activities. The technology evolution has brought the ion channel research and drug discovery to a new era, which will be discussed in later part of this chapter.

2.1. Genetic evolution and expression

Compared to Cav and Kv channel family, the rise of Nav family is relatively recent [18]. The study of intron/exon organization suggested that Navs were evolved from the similarly structured Cav channels. This is supported by the finding that the four domains of Navs have higher similarity to their corresponding domains in the Cav channels than to each other. The ancestral Navs and Cavs genes might have evolved by two rounds of gene duplication, i.e., from an ancestral, single-domain Kv-like gene to a two-domain protein, then from a two-domain protein to a four-domain protein. This hypothesis is also in line with the observation that among the four domians of Navs, DI shares higher similarity with DIII, and DII shares higher similarity to DIV [19].

In choanoflagellate (the sister group of animals), the rapid long-distance communication among excitable cells is achieved at the emergence of Metazoa (represented by bilaterian animals and cnidarians) through the use of Nav channel [20, 21]. The gene organization, biophysical, and pharmacological properties of invertebrate sodium channels are largely similar to their mammalian counterparts, suggesting that the primordial Nav channels were established before the evolutionary separation of the invertebrates from the vertebrates, and evolution of Navs played a critical role in the emergence of nervous systems in animals [19, 22]. Of note, sodium selectivity might be acquired independently in BacNav and mammalian Nav channels as indicated by phylogenetic analysis [23]. Therefore, BacNav channels can serve as models for studying Navs structure function, but evolutionary variation should be taken into consideration.

Historically, the tissue distribution of mammalian Nav isoforms was obtained by methods such as quantitative PCR, expressed sequence tag (EST) profiling, and pharmacology study using isoform selective toxins. The more precise expression data were recently obtained by microarray [24, 25] and mRNA sequencing from Genotype-Tissue Expression (GTEx) project [26] (Figure 5). Now, we know that Nav1.1, Nav1.2, and Nav1.3 are predominantly expressed in central nervous system (CNS). Nav1.1 is the predominant channel in the caudal regions and the spinal cord, though relatively low-level expression of Nav1.1 has been shown in the peripheral nervous system (PNS). Nav1.2 is the highest expression isoform in the rostral regions. Nav1.3 peaks at birth but remains detectable at a lower level in adulthood. Interestingly, all of these CNS-enriched isoforms are sensitive to tetrodotoxin (TTX) at nanomolar concentrations, and their genes are clustered on chromosome 2 in both mice and humans.

Nav1.4 and Nav1.5 exhibit strikingly high-level expression in muscle and heart, respectively. More sensitive approaches have detected Nav1.5 in the piriform cortex and limbic regions of the brain but at relatively low level [27]. These two Nav isoforms can be distinguished from each other and from the CNS isoforms on the basis of toxin sensitivity. Adult skeletal muscle-enriched Nav1.4 is sensitive to TTX and μ conotoxin GIIIA at nanomolar concentrations, while the CNS-enriched channels are only sensitive to TTX, and heart-specific Nav1.5 is resistant to TTX. In addition, Nav1.4 is inhibited by nanomolar concentrations of μ conotoxin PIIIA, whereas Nav1.2 is approximately 15-fold less sensitive, and Nav1.7 is resistant [28, 29, 30].

Nav1.6 appears to be abundantly expressed in both PNS and CNS tissues. Nav1.7, Nav1.8, and Nav1.9 are primarily expressed in PNS, including nociceptive neurons, Aβ-fibers, C-fibers, and both large and small diameter dorsal root ganglion (DRG) [31]. This PNS-specific localization of Nav1.7, Nav1.8, and Nav1.9 make them attractive targets for developing isoform selective modulators to treat many PNS-related diseases and pathologic conditions.

2.2. Nav channel structure-function

The core functional unit of Nav channels is α subunit. Each α subtype is composed of a single polypeptide chain with ∼2000 amino acid residues forming four pseudoheteromeric domains designated as DI to DIV. Nine α subunits’ protein sequence homology is greater than 70% (Table 1). The structure complexity and high-sequence homology make it difficult to design subtype-selective drugs. In 2011, the first X-ray crystal structure of bacterial Arcobacter butzleri Nav channels (NavAb) was determined [32]. Subsequently, structures for several BacNavs and a BacNav-human Nav chimeric channel have been resolved, representing closed [33, 34], open [34, 35], and potentially inactive states of the channels [36, 37]. In 2017, the first two eukaryotic Nav structures for American cockroach and electric eel (Ee) Nav1.4-β1 complex were determined by using cryogenic electron microscopy (CryoEM) [12]. Findings from crystal and cryo-EM structures are mostly consistent, and collectively provided important insights into Nav channel structure-function and structure-based drug design.

Voltage-sensing domains (VSDs). The S1 to S4 segments form voltage-sensing domain. There are four VSDs in a sodium channel. Each VSD is featured by repetitively occurring positively charged Lys or Arg residues at every third position in S4. These charge clusters move toward the extracellular surface upon membrane depolarization and return to their resting positions upon membrane repolarization. Each VSD is connected to an intracellular S4-S5 linker, which transfers the movement of S4 segment to the central pore domain formed by S5-S6. Thus, the outward and inward movements of S4 result in channel opening and closing, respectively. Unlike homotetrameric Kvs and BacNavs, mammalian Nav channel’s four VSDs possess distinctive sequence signature, conformation, and role in channel gating [12, 13]. For examples, the four VSDs have nonconserved intra- and extracellular loops, and the different charge clusters in S4, i.e., RRRR in DI, RRRRK in DII, KRRRR in DIII, and RRRRR in DIV. Also, the S2 in each VSD also makes asymmetric functional contributions to Nav channel activation and inactivation [38].

Pore domain (PD): the S5 to S6 segments from the four domains of the α subunits enclose the central pore of the channel. The extracellular linker connecting S5-S6 is defined as pore-loop (P-loop), which can be separated into two α helices named P1 and P2. The functional entities along the ion permeation pathway in PD include the selectivity filter (SF), the central cavity, and the intracellular activation gate, as shown in Figure 4E.

The outer vestibule and selectivity filter are formed in P-loop reentering membrane segments, designated as P1-SF-P2 funnel. Early study by comparing Navs with Cavs found that one residue, Asp/Glu/Lys/Ala (DEKA), at the corresponding locus in the middle of P1-P2 determines Na⁺ selectivity [39]. Structure studies confirmed that the asymmetric selectivity filter vestibule is constituted by the side chains of the signature DEKA residues and the carbonyl oxygens atoms of the two preceding residues in each domain, Thr/Gln (DI), Cys/Gly (DII), Thr/Phe (DIII), and The/Ser (DIV). Mutational study identified an additional outer ring above the selectivity filter, Glu/Glu/Met/Asp (EEMD), which significantly interfered the tetrodotoxin (TTX) binding [40, 41]. The outer vestibule and SF structures were further discerned by using bacteria KcsA channel X-ray structure as template and guanidinium toxins (TTX and saxitoxin, STX), which successfully defined the first pharmacological relevant site on Nav channels, site 1 [42]. After that, local anesthetic binding site was determined within the four fenestrations in PD, each with distinct shape and size [13, 43]. From studying a group of activators, including batrachotoxin (BTX), veratridine (VTD), grayanotoxin (GTX), and aconitine (ACD), the neurotoxin site 2 was determined in the inner cavity of PD [44, 45].

Activation gate: the activation gate of Nav channels was originally predicted to be at the inner end of the pore based on the study of local anesthetics, which exhibit usage-dependent blockage [46, 47]. From thiol-modifying reagent accessibility study, a ∼3.8 Å diameter constriction formed by a ring of conserved hydrophobic residues at the end of S6 (represented as DI-V440, DII-L795, DIII-I1287, DIV-I1590 in Nav1.4 and DI-Y405, DII-F960, DIII-F1449, DIV-F1752 in Nav1.7) was found to occlude only the pore at closed state but not at open state [48]. Recent cryo-EM structures confirmed that this activation gate is located at the cytoplasmic boundary level of the membrane [12]. Channelopathy study found that Nav1.7 DII S6 L955 deletion cause F960 side change conformational change in the activation gate. This mutation produces radial shift of the channel open at 25 mV more hyperpolarizing voltage, which renders hyperexcitability of DRG neurons to cause inherited erythromelalgia (IEM) [48].

Fast inactivation: while the channel opening is controlled by VSD and activation gate, the fast inactivation is regulated by a highly conserved Ile/Phe/Met (IFM) motif, which is localized in an intracellular loop connecting the domain DIII and DIV [49]. The IFM motif was originally discovered as a particle segment attached to the inner end of the pore. Study showed that pronase and N-bromoacetamide removed inactivation only when applied from intracellular side [50], and acetyl-KIFMK-amide peptides restored fast inactivation [51]. In the eukaryotic EeNav1.4-β1 structure at open and resting states, the LFM motif (equivalent to IFM) in DIII-IV linker is plugged into the corner enclosed by the outer S4-S5 and inner S6 segments in DIII and DIV. Once the channel opens, the LFM motif acts as a hinged lid and folds into the intracellular mouth of the open pore to produce fast inactivation.

Slow inactivation: in contrast to fast inactivation (milliseconds scale), slow inactivation occurs in seconds during prolonged depolarization or rapid repetitive stimulations. Slow inactivation determines channel availability for action potential generation; thus, it endows neuronal tissues with memory of previous excitation, prevents excitation of skeletal muscle by mild hyperkalemia, and affects the conduction velocity and excitability of cardiac tissue. For the cardiac Nav1.5 channel, two putative proton sensor residuals in P-loop, C373 and H880, were responsible of tissue-acidification-induced slow inactivation underlying cardiac arrhythmia [52]. Other mutations, including DII S4-S5 linker L689I in Nav1.4, DII S6 Del-L955 in Nav1.7, have also been identified to impair slow inactivation, causing hyperkalemic periodic paralysis and inherited erythromelalgia, respectively [53, 54]. Voltage-clamp fluorimetry (VCF) study of Nav1.4 channels showed that that immobilization of DI and DII VSDs is involved in the development of slow inactivation, while DIII VSD is involved in the recovery from slow inactivation [55]. However, the structural basis for slow inactivation remains undefined.

2.3. Nav β subunits

In vertebrates, five Nav auxiliary β subunits, β1, β1B, β2, β3, and β4, have been identified (SCN1B to SCN4B), with molecular weight ranging from 30 to 40 kD (Table 1) [57, 58, 59, 60]. In invertebrates, auxiliary subunits such as tipE and Vssc β in drosophila bear no homology to vertebrate β subunits, suggesting a separate evolutionary pathway [61, 62, 63]. All β subunits comprise an amino terminal immunoglobulin (Ig) domain, a single transmembrane (TM) segment, and an intracellular carboxyl-terminal, except for β 1B (previously named β1A), which is a β1 splice variant that lacks TM segment [64]. β1 and β3 interact with α subunit noncovalently, whereas β2 and β4 bind to the α subunit via a disulfide bond. All β isoforms are expressed in CNS, PNS, and cardiovascular systems, including excitable and nonexcitable cells, where they are part of the V-set immunoglobulin superfamily of cell adhesion molecules facilitating cell adhesion and cell migration. β1 is highly expressed in skeletal and cardiac muscles, and the expression patterns of all β isoforms vary during development [65]. β subunits play broad role in modulating Nav function. They regulate expression and membrane trafficking of α subunits, modulate channel activation and inactivation, and interfere with toxin binding [66, 67, 68].

In 2017, the cryo-EM strucutre for EeNav1.4 in complex with the β1 subunit was determined [12]. This structure provided a first glimpse into the interaction between α and β subunits. The β1 subunit interacts with the α subunits as an ax, wherein its TM interacts with VSDIII within the membrane as the handle, while its Ig domain as the head interacts with DI-L5 and DIV-L6 extracellular loops and the intervening segment between DIII S1 and S2 (Figure 4C and D). Mutations in β subunits have been linked to many human diseases, including epilepsy, and cardiac arrhythmia, and sudden death syndromes. Although β subunit-specific drugs have not yet been developed, the Nav β subunit family remains a potential therapeutic target [68].

3.1. Pain

Nav1.7, 1.8, and 1.9 are highly expressed in sensory neurons. They control the excitability of nociceptive neurons, and thus are considered as therapeutic targets for pain relief [31, 71, 72, 73, 74]. Among them, Nav1.7 was the first gene linked to human pain. In 2004, two missense mutations, I848T and L858H in SCN9A (Nav1.7), were associated with edema, redness, warmth, and bilateral pain in human inherited erythromelalgia (IEM) patients [75]. In 2006, three nonsense mutations, S459X, I767X, and W897X, were identified in congenital insensitivity to pain (CIP) patients [76]. Since then, additional gain-of-function mutations are associated with IEM, paroxysmal extreme pain disorder (PEPD), small fiber neuropathy (SFN), and additional loss-of-function mutations that are associated with CIP (Figure 6). The mechanism underlying these conditions was unraveled by characterizing biophysical properties of disease mutations [76] and nociceptor-specific Nav1.7 knockout [77]. Together, these studies have established that Nav1.7 is an essential and nonredundant requirement for nociception in humans.

While Nav1.7 is responsible for setting threshold for generation of action potentials, Nav1.8 and Nav1.9 contribute to the rising phase of action potentials in nociceptive neurons. Both channels are expressed in small-diameter DRG neurons, which include the C fibers that transmit nociceptive impulses [78]. Four gain-of-function Nav1.8 mutations (L554P, I1706V, A1304T, G1662S) have been identified, which lead to an increase in excitability in small-diameter neurons, underlying pain in small fiber neuropathy (SFN) [79, 80, 81]. In Nav1.9, gain-of-function mutations were recently reported to be associated with pain, albeit with opposing effects [82]. So far, naturally occurring loss-of-function mutations of Nav1.8 and Nav1.9 are yet to be described in humans; Nav1.8 and 1.9 are clearly important in the pain pathology and worth exploring as potential pain targets.

3.2. Epilepsy

Nav channel dysfunction is central to the pathophysiology of epileptic seizures, and many of the most widely used antiepileptic drugs, including phenytoin, carbamazepine, and lamotrigine, are Nav inhibitors. Many types of general epilepsy are resulted from mutations primarily in Nav1.1, and also in other Nav α isoforms, including Nav1.2, Nav1.3, Nav1.6, Nav1.7, and Nav β1. GEFS+ type 1 results from Nav β1 mutation C387G, which destroys extracellular immunoglobulin domain and thus indirectly decreases rate of channel inactivation [83]. GEFS+ type 2 results from Nav1.1 mutations, such as T875 M (in DII-S4) and R1648H (in DIV-S4), which decrease Nav1.1 inactivation rate directly [84]. Worth noting, some epilepsy-assicated Nav1.2 and Nav1.6 mutations cause a gain-of-function when characterized in transfected cells [85, 86]. Better understanding of clinical genetics and channel structure-function will facilitate the drug development for Nav-associated neurological diseases.

3.3. Cardiac arrhythmias

Mutations of Nav1.5 have been linked to long QT syndrome (LQTS). Recent study suggests that even modest depression of Nav1.5 expression may promote pathologic cardiac remodeling and progression of heart failure [87]. Since the first LQT-related Nav1.5 mutation being discovered in 1995 [88], more than a hundred Nav1.5 mutations that associate with distinct cardiac rhythm disorders, such as LQT syndrome subtype 3, Brugada syndrome, and cardiac conduction disease, have been identified [89]. These Nav1.5 mutations are spread out the whole protein. Most Nav1.5 mutations change biophysical property by increasing persistent Na⁺ current or gain-of-function, while in Brugada syndrome and cardiac conduction disease, most of the mutations are missense loss-of-function mutations. Because many mutations produce overlapping clinical phenotypes, it is crucial to understand genotype-phenotype correlations in Nav1.5 channelopathies for drug development.

3.4. Neuromuscular diseases

Mutations of Nav1.4 channel are associated with inherited neuromuscular diseases. For examples, the M1592V mutation causes hyperkalemic periodic paralysis (HYPP), in which increased levels of serum potassium lead to muscle hypoexcitability and paralysis; the R1448C mutation causes paramyotonia congenita (PMC), which is induced by cold and aggravated by increased muscle activity; and the G1306A mutation causes potassium-aggravated myotonias (PAM). These mutations affect either voltage-dependent activation or inactivation of Nav1.4, and are inherited in an autosomal dominant manner [90]. In contrast, most Nav1.6 mutations are recessive. An allelic mutation A107T in DIII S4-S5 caused ataxic phenotype by shifting Nav1.6 activation and inactivation to about 14 mV in the depolarizing direction, suggesting Nav1.6 is required for the complex spiking of cerebellar Purkinje cells and for persistent sodium current in several classes of neurons [91, 92, 93].