Chapter 6:  Electrical Excitability and Ion Channels

All excitable cells have a membrane potential

At rest, the entire cytoplasm is electrically more negative than the external bathing fluid by 30 to 100 mV. All of this potential drop appears across the extremely thin external cell membrane, as may be ascertained by recording with an electrolyte-filled glass pipette microelectrode. When such an electrode is used to probe potentials around an excitable cell, a sudden negative drop appears at the moment the narrow tip of the pipette breaks through the cell surface. By convention, the membrane potential is always reported in terms of “inside” minus “outside,” so the resting potential is a negative number, for example, −70 mV in a myelinated nerve fiber. Signals that make the cytoplasm more positive are said to depolarize the membrane, and those making it more negative are said to hyperpolarize the membrane.

Electrical signals recorded from cells are basically of two types: stereotyped action potentials characteristic of each cell type and a variety of slow potentials

The action potential of axons is a brief, spike-like depolarization that propagates regeneratively as an electrical wave without decrement and at a high, constant velocity from one end of the axon to the other [3]. It is used for all rapid signaling over a distance. For example, in the reflex arc of Figure 6-1, action potentials in motor axons might carry the message from spinal cord to arm, telling the muscle fibers of the biceps to contract. In large mammalian axons at body temperature, the action potential at any one patch of membrane may last only 0.4 msec as it propagates at a speed of 100 m/sec. The action potential normally is elicited when the cell membrane is depolarized by some type of stimulus to beyond a threshold level; it is said to be produced in an all-or-nothing manner because a subthreshold stimulus gives no propagated response, whereas every suprathreshold stimulus elicits the stereotyped propagating wave. Underlying the propagated action potential is a regenerative wave of opening and closing of voltage-gated ion channels that sweeps along the axon. Action potentials also are frequently referred to as spikes or impulses. Most nerve cell and muscle cell membranes can make action potentials and are said to be electrically excitable, but a few cannot.

By contrast, slow potentials are localized membrane depolarizations and hyperpolarizations, with time courses ranging from several milliseconds to minutes. They are associated with a variety of transduction mechanisms. For example, slow potentials arise at the synaptic site of action of neurotransmitter molecules (see Chap. 10) and of some hormones, as well as in sensory endings of chemosensors, mechanosensors and photoreceptors (see Chaps. 47 and 48). These electrical signals in sensory endings frequently are called generator or receptor potentials, and the signals arising postsynaptically at chemical synapses are called postsynaptic potentials. Slow potentials are graded in relation to their stimulus and sum with each other both spatially and temporally within the cell, while decaying passively over an intracellular distance of no more than a few millimeters from the site of generation. Underlying the slow potential is a graded and local opening or closing of ion channels, reflecting the intensity of the stimulus. These channels are gated by stimuli other than voltage. The natural stimulus for the initiation of propagated action potentials is a depolarizing slow potential exceeding the firing threshold. For example, impulses in a wide variety of presynaptic cells often give rise to a barrage of excitatory, or depolarizing, and inhibitory, or hyperpolarizing, postsynaptic potentials in the dendrites of a postsynaptic neuron. These slow potentials sum within the dendrites and cell body to provide the drive stimulating or suppressing initiation of action potentials at the axon hillock, which then propagate down the axon. In each of the four cell populations involved in the reflex arc in Figure 6-1, depolarizing slow potentials give rise to propagating action potentials as the message moves forward.

How do membrane potentials arise?

Consider the electrolyte system represented in Figure 6-2 (left), where a porous membrane separates aqueous solutions of unequal concentrations of a fictitious salt, KA. Two electrodes permit the potential difference between the two solutions to be measured. Now, assume that the membrane pores are permeable exclusively to K⁺ so that K⁺ begins to diffuse across the membrane but A⁻ does not. For simple statistical reasons, the movement of K⁺ from the concentrated side to the dilute side initially will exceed the movement in the reverse direction, and we expect a net flux of K⁺ down its concentration gradient. However, this process does not continue long since K⁺ carries a positive charge from one compartment to the other and leaves a net negative charge behind. The growing separation of charge creates an electrical potential difference, the membrane potential, between the two solutions; and a positive charge appears on the side into which the K⁺ ions diffuse, thereby setting up an electrical force that tends to oppose further net movement of K⁺.

Equilibrium potential is the membrane potential at which there are no net ion movements

The membrane potential reached in a system with only one permeant ion and no perturbing forces is the equilibrium, or Nernst, potential for that ion; thus, the final membrane potential for the system in Figure 6-2 is the potassium equilibrium potential E_K. At that potential, there is no further net movement of K⁺ and, unless otherwise disturbed, the membrane potential and ion gradient will remain stable indefinitely. The value of the Nernst potential is derived from thermodynamics by recognizing that the change of electrochemical potential, Δ μ_j, for moving the permeant ion j^+z across the membrane must be zero at equilibrium: (1) where R is the gas constant (8.31 J/degree/mol), T is absolute temperature in Kelvin (°C + 273.2) and F is Faraday's constant (96,500 C/mol). Using terms appropriate to biology, [j]_o and [j]_i represent activities of ion j^+z outside and inside a cell, z is the ionic valence and E is the membrane potential defined as “inside minus outside.” Solving for E and calling it E_j to denote the ion at equilibrium gives the Nernst equation for ion j:

(2)

For practical use at 20°C, the Nernst equation can be rewritten: (3) showing that for a 10:1 transmembrane gradient, a monovalent ion can give 58 mV of membrane potential. Table 6-1 gives approximate intracellular and extracellular concentrations of the four electrically most important ions in a mammalian skeletal muscle cell and the Nernst potentials calculated from these numbers at 37°C, neglecting possible activity coefficient corrections. Experimentally, it is found that the resting muscle membrane is primarily permeable to K⁺ and Cl⁻, and therefore, the resting potential in muscle is −90 mV, close to the equilibrium potentials E_K and E_Cl. During a propagated action potential, ion channels permeable to Na⁺ open, some Na⁺ enters the muscle fiber and the membrane potential swings transiently toward E_Na. When these pores close again, the membrane potential returns to near E_K and E_Cl. To summarize, membrane potentials arise by diffusion of a small number of ions down their concentration gradient across a permselective membrane.

Real cells are not at equilibrium

Although the concept of equilibrium potentials is essential to understand and predict the membrane potentials generated by ion permeability, real cells are never at equilibrium because different ion channels open and close during excitation and even at rest several types of channel are open simultaneously. Under these circumstances, the ion gradients are dissipated constantly, albeit slowly, and ion pumps are always needed in the long run to maintain a steady state (see Chap. 5). The net passive flux, M_j, of each ion is proportional to the permeability, P_j, for that ion and often is given, at least approximately, by an empirical formula called the Goldman-Hodgkin-Katz flux equation [3–5]:

(4)

Experimentally, these fluxes may be measured as an electrical current or by using radioactive tracers or with sensitive indicator substances responding to the ion in question by fluorescence or other optical changes. In most cases, the fluxes are too small to detect by the less sensitive classical method of chemical analysis for the total amount of an ion.

When the membrane is permeable to several ions, the steady-state potential is given by the sum of contributions of the permeant ions, weighted according to their relative permeabilities:

(5)

The Goldman-Hodgkin-Katz voltage equation often is used to determine the relative permeabilities to ions from experiments where the bathing ion concentrations are varied and changes in the membrane potential are recorded. It has the same form as the equation usually used to describe the responses of ion-selective electrodes in analytical work in the laboratory.

During excitation, ion channels open or close, ions move and the membrane potential changes

The extra ion fluxes during activity act as an extra load on the Na⁺-K⁺ and the Ca²⁺ pumps, consuming ATP and stimulating an extra burst of cellular oxygen consumption until the original gradients are restored (see Chap. 5). How large are these fluxes? The physical minimum, calculated from the rules of electricity, is a very small number. Only 10⁻¹² equivalents of charge need be moved to polarize 1 cm² of membrane by 100 mV, meaning that ideally the movement of 1 pmol/cm² of monovalent ion would be enough to depolarize the membrane fully. This quantity, related to the electrical capacitance of the membrane, is a constant throughout the animal and plant kingdoms, as would be the case if the effective thickness and dielectric constant of the hydrophobic, insulating part of all cell plasma membranes were similar. In practice, unmyelinated axons gain about 4 to 8 pmol of Na⁺ and lose about the same amount of K⁺ per square centimeter for one action potential. The figure is higher than the physical ideal because the oppositely directed fluxes of Na⁺ and K⁺ overlap considerably in time, working against each other. With this kind of Na⁺ gain, an unmyelinated squid giant axon of 1-mm diameter could be stimulated 10⁵ times and a mammalian fiber of 0.2-μm diameter only 10 to 15 times before the internal Na⁺ concentration would be doubled, assuming that the Na⁺-K⁺ pump had been blocked. In myelinated nerve, the Na⁺ gain in one impulse is very small, amounting to only 2 × 10⁻⁷ mol/kg of nerve because of the special low-capacitance properties of myelin.

Transport systems also may produce membrane potentials

The equations just discussed are those for passive electrodiffusion in ion channels, where the only motive forces on ions are thermal and electrical, and they do indeed explain almost all of the potentials of excitable cells. However, there is another type of electrical current source in cells that can generate potentials: the ion pumps and other membrane devices that couple ion movements to the movements of other molecules. In excitable cells, the most prominent is the Na⁺-K⁺ pump (see Chap. 5), which gives a net export of positive charge and, hence, tends to hyperpolarize the cell surface membrane in proportion to the rate of pumping [3]; but hyperpolarization from this electrogenic pumping is typically only a modest few millivolts. By contrast, mitochondria, as well as plant, algal and fungal cells, have powerful current sources in their proton-transport system. Their membrane potentials often are dominated by this electrogenic system and, thus, are not describable in terms of diffusion in simple passive channels.

Permeability changes of the action potential

Given the rules of ionic electricity, the major biological problem in understanding action potentials is to describe and explain the ion permeability mechanisms in the membrane. The opening and closing of ion channels involves conformational changes driven by electrical field changes or ligand binding but not by direct consumption of metabolic energy. The independence of immediate metabolic input can be demonstrated in studies with internally perfused cells and with channels reconstituted into lipid bilayers. For example, the great majority of the axoplasm can be squeezed from one cut end of a squid giant axon and the axon reinflated with a continuously flowing salt solution that enters at one end and leaves at the other, and the axon can continue to fire >100,000 impulses. Analogous experiments using dialysis techniques or excised patches of membrane have been done with many other excitable cells. These experiments prove that ATP and other intracellular, small molecules of metabolism are not required either for many cycles of opening and closing of Na⁺, K⁺ or Ca²⁺ channels or for the resulting depolarizing and repolarizing ionic current flows. They also show that intracellular ATP, cGMP and Ca²⁺, as well as phosphorylation by a variety of protein kinases, can be powerful modulators of channel activities. In the long term, ATP and other molecules also are needed to fuel the Na⁺-K⁺ and Ca²⁺ pumps and for synthesis and trafficking of membrane components. We must emphasize that channels differ from pumps (see Chap. 5) in their structure, mechanism of ion flux, function and regulation.

Gating mechanisms for Na⁺ and K⁺ channels in the axolemma are voltage dependent

In a classic series of experiments, Hodgkin, Huxley, and Katz [1,3–6] measured the kinetics of ion permeability changes in squid giant-axon membranes by a direct electrical method called the voltage clamp. The method controls the membrane voltage electrically, usually with step changes of potential, while ion movements are recorded directly as electrical current flowing across the membrane. The recorded current may be resolved into individual ionic components by changing the ions in the solutions that bathe the membrane. The voltage clamp is a rapid and sensitive assay for studying the opening and closing of ion channels. A widely used miniature version of the voltage clamp is the patch clamp, a technique with sufficient sensitivity to study the current flow in a single ion channel [7]. A glass micropipette with a tip diameter <1 μm is fire-polished at the tip and then pressed against the membrane of a cell. Because the tip is smooth, it seals to the membrane in the annular contact zone, rather than piercing the membrane, and defines a tiny patch of the cell surface whose few ion channels can be detected easily by the currents flowing through them. The patch clamp can readily measure a flux of as little as 10⁻²⁰ mol of ion in less than 1 msec.

With the voltage clamp, Hodgkin and Huxley [6] discovered that the processes underlying gating, that is, the opening and closing conformational changes, of axonal Na⁺ and K⁺ channels are controlled by the membrane potential and, therefore, derive their energy from the work done by the electrical field on charges associated with the channel macromolecule. Hodgkin and Huxley [6] identified currents from two types of ion-selective channel, Na⁺ and K⁺ channels, which account for almost all of the current in axon membranes; and they made a kinetic model of the opening and closing steps, which may be simplified as shown in Figure 6-3. Depolarization of the membrane is sensed by the voltage sensor of each channel and causes the conformational reactions to proceed to the right. Repolarization or hyperpolarization causes them to proceed to the left. We can understand the action potential in these terms. The action potential, caused by a depolarizing stimulus, begins with a transient, voltage-gated opening of Na⁺ channels that allows Na⁺ to enter the fiber and depolarize the membrane fully, followed by a transient, voltage-gated opening of K⁺ channels that allows K⁺ to leave and repolarize the membrane. Figure 6-4 shows a calculation of the temporal relation between channel-opening and membrane-potential changes in an axon at 18.5°C, using the model of Hodgkin and Huxley [6].

The action potential is propagated by local spread of depolarization

If there are no chemical or mechanical signals for voltage-gated channels to open, how does the action potential propagate smoothly down an axon, bringing new channels into play ahead of it? Any electrical depolarization or hyperpolarization of a cell membrane spreads a small distance in either direction inside an axon from its source by a purely passive process often called cable, or electrotonic, spread. The spread occurs because the intracellular and extracellular media are much better conductors than the membrane, so any charges injected at one point across the membrane repel each other and disperse along the membrane surface. The lower part of Figure 6-4 shows diagrammatically the so-called local circuit currents that spread the depolarization forward. In this way, an excited depolarized membrane area smoothly depolarizes the next unexcited region ahead of the action potential, bringing it above firing threshold, opening Na⁺ channels there and advancing the wave of excitation. The action potential in the upper part of Figure 6-4 is calculated by combining the known geometry of the squid giant axon with the rules of ionic electricity and the kinetic Hodgkin-Huxley equations for the voltage-dependent gating of Na⁺ and K⁺ channels. The success of the calculations means that the factors described are sufficient to account for action-potential propagation.

Membranes at nodes of Ranvier are characterized by high concentrations of Na⁺ channels

A wide variety of cells have been studied by voltage clamp methods, and quantitative descriptions of their permeability changes are available. All axons, whether vertebrate or invertebrate, operate on the same principles: they have a small background permeability, primarily to K⁺, which sets the resting potential and display brief, dramatic openings of Na⁺ and K⁺ channels in sequence to shape the action potentials. Chapter 4 describes myelin, a special adaptation of large (1- to 20-μm diameter) vertebrate nerve fibers for higher conduction speed. In myelinated nerves, like unmyelinated ones, the depolarization spreads from one excitable membrane patch to another by local circuit currents; but because of the insulating properties of the coating myelin, the excitable patches of axon membrane, the nodes of Ranvier, may be more than 1 mm apart, so the rate of progression of the impulse is faster. Nodes of Ranvier have Na⁺ channels similar to those of other axon membranes, but nodal membranes have at least ten times as many channels per unit area to depolarize the long, passive, internodal myelin. The Na⁺-K⁺ pump may be distributed similarly (see Chap. 5). The internodal axon membrane has K⁺ channels but far fewer Na⁺ channels. After experimental demyelination by diphtheria toxin, which takes several days, and probably in the course of several demyelinating diseases (see Chap. 39), Na⁺ channels and excitability can develop in a formerly unexcitable internodal section of axon.

A wide repertoire of voltage-sensitive channels is found among cell types

A diversity of channel types is found in the different cell types in any one organism, where the repertoire of functioning channels is adapted to the special role each cell plays in the body. We know of more than 50 genes for the pore-forming subunits of channels in the voltage-gated family. In many cell types, it is not uncommon to find several Ca²⁺ channels that open with depolarization, supplementing the depolarizing effect of Na⁺ channels by adding a slower depolarizing Ca²⁺ influx or sometimes even acting alone to depolarize the membrane without Na⁺ channels [8]. Ca²⁺ channels have a special importance because the entering Ca²⁺ often plays the role of a chemical messenger to activate exocytosis, or secretion; contraction; gating of other channels; ciliary reorientation; metabolic pathways; gene expression; and other processes. Indeed, whenever an electrical message activates any nonelectrical event, a change of the intracellular free Ca²⁺ concentration acts as an intermediary. Ca²⁺ channels are particularly concentrated in nerve terminals, where a Ca²⁺ influx is required for release of chemical neurotransmitters.

Ion channels are macromolecular complexes that form aqueous pores in the lipid membrane

We have learned much about ion channel function from voltage clamp and patch clamp studies on channels still embedded in the cell membrane [1–5]. Figure 6-5A summarizes the major functional properties of a voltage-gated macromolecular channel in terms of a fanciful cartoon. The pore is narrow enough in one place, the ionic selectivity filter, to “feel” each ion and to distinguish among Na⁺, K⁺, Ca²⁺ and Cl⁻. The channel also contains charged components that sense the electrical field in the membrane and drive conformational changes that, in effect, open and close gates controlling the permeability of the pore. In Na⁺, K⁺ and Ca²⁺ channels, the gates seem to close the axoplasmic mouth of the pore and the selectivity filter seems to be near the outer end of the pore.

How do we know that a channel is a pore? By far the most convincing evidence is the large ion flux a single channel can handle. It is not unusual in patch clamp work to measure ionic currents of 2 to 10 pA flowing each time one channel in the patch is open. This would correspond to 12 to 60 × 10⁶ monovalent ions moving per second. Such a turnover is several orders of magnitude faster than known carrier mechanisms and agrees well with the theoretical properties of a pore of atomic dimensions. Similar fluxes have been observed with pore-forming antibiotic peptides in model systems, including gramicidin A, alamethicin and monazomycin.

Water molecules break and make hydrogen bonds with other waters 10¹¹ to 10¹² times per second, and alkali ions exchange water molecules or other oxygen ligands at least 10⁹ times per second. In these terms, the progress of an ion across the membrane is not the movement of a fixed hydrated complex; rather, it is a continual exchange of oxygen ligands as the ion dances through the sea of relatively free water molecules and polar groups that form the wall of the pore. It is generally assumed that polar and charged groups are in the pore to provide stabilization energy to the permeating ion, compensating for those water molecules that must be left behind as the ion enters into the pore. Evidence for important negative charges in the selectivity filter of Na⁺, K⁺ and Ca²⁺ channels comes from a block of their permeability as the pH of the external medium is lowered below 5.5 [5] and from site-directed mutagenesis of aspartate and glutamate residues in cloned channels (see below).

The minimum size of ion channels has been determined from the van der Waals dimensions of ions that will go through them [5]. Voltage-gated channels with considerable ion selectivity seem to be so narrow that ions need to shed several, though not all, water molecules to pass through. The ion fluxes often are described by models with temporary binding to attractive sites and jumps over energy barriers. Formally, the kinetics of flux through channels are construed similarly to enzyme kinetics. It is assumed that the channel passes through a sequence of “channel—ion complexes” as it catalyzes the progression of an ion across the membrane. Such theories also can describe other properties of ion channels, such as selectivity, saturation, competition and block by permeant ions [5].

Voltage-dependent gating requires voltage-dependent conformational changes in the protein component(s) of ion channels

On theoretical grounds, a membrane protein that responds to a change in membrane potential must have charged or dipolar amino acid residues, located within the membrane electrical field and acting as voltage sensors, as illustrated in Figure 6-5A. Changes in the membrane potential then exert a force on these protein-bound charges. If the energy of the field—charge interactions is great enough, the protein may be induced to undergo a change to a new stable conformational state in which the net charge or the location of charge within the membrane electrical field has been altered. For such a voltage-driven change of state, the steepness of the state function versus membrane potential curve defines the equivalent number of charges that move according to a Boltzmann distribution. On this basis, activation of Na⁺ channels would require the movement of six to 12 positive charges from the intracellular to the extracellular side of the membrane. The movement of a larger number of charges through a proportionately smaller fraction of the membrane electrical field would be equivalent. Good candidates for such gating charges have been identified in the amino acid sequences of voltage-gated channels.

Such movements of membrane-bound charge give rise to tiny “gating” currents, which can be detected electrophysiologically [9]. Their voltage and time dependence are consistent with the multistep changes of channel state from resting to active. In contrast to activation, fast inactivation from the open state of Na⁺ channels and certain K⁺ channels does not seem to be a strongly voltage-sensitive process. This inactivation can be blocked irreversibly by proteolytic enzymes acting from the intracellular side of the channel. Regions of ion channels that are exposed at the intracellular surface of the membrane are important in mediating the process of inactivation.

Pharmacological agents acting on ion channels help define their functions

The Na⁺ channel is so essential to successful body function that it has become the target in the evolution of several potent poisons. The pharmacology of such agents has provided important insights to the further definition of functional regions of the channel [2,4,5]. Figure 6-5B shows the supposed sites of action of four prominent classes of Na⁺ channel agents. At the outer end of the channel is a site where the pufferfish poison, tetrodotoxin (TTX), a small lipid-insoluble charged molecule, binds with a K_i of 1 to 10 nM and blocks Na⁺ permeability. A second important class of Na⁺ channel blockers includes such clinically useful local anesthetics as lidocaine and procaine and related antiarrhythmic agents. They are lipid-soluble amines with a hydrophobic end and a polar end, and they bind to a hydrophobic site on the channel protein, where they also interact with the inactivation-gating machinery. The relevant clinical actions of local anesthetics are explained fully by their mode of blocking Na⁺ channels. Two other classes of toxins either open Na⁺ channels spontaneously or prevent them from closing normally once they have opened. These are lipid-soluble steroids, such as the frog-skin poison, batrachotoxin (BTX); the plant alkaloids, aconitine and veratridine, both acting at a site within the membrane; and peptide toxins from scorpion and anemone venoms, which act at two sites on the outer surface of the membrane. Most scorpion and anemone toxins specifically block the inactivation-gating step. It is interesting to note that the affinity of the channel for each of these classes of toxins depends on the gating conformational state of the channel.

Similarly specific agents affect K⁺ and Ca²⁺ channels. Most K⁺ channels can be blocked by tetraethylammonium ions, by Cs²⁺ and Ba²⁺ and by 4-aminopyridine. Except for 4-aminopyridine, there is good evidence that these ions become lodged within the channel at a narrow place, from which they may be dislodged by K⁺ coming from the other side [2]. In addition, certain K⁺ channels can be distinguished by their ability to be blocked by polypeptide toxins, such as charybdotoxin from scorpion, apamin from bee or dendrotoxin from snake. Ca²⁺ channels can be blocked by externally applied divalent ions, including Mn²⁺, Co²⁺, Cd²⁺ and Ni²⁺. Different Ca²⁺ channel subtypes can be distinguished by their block by the dihydropyridines, nifedipine, ω-conotoxins from the cone snail and agatoxins from spider [8].

Radiolabeled neurotoxins that act on Na⁺ channels are used as molecular probes to tag the channel proteins, allowing their identification

Neurotoxins act at several different sites on Na⁺ channels to modify their properties (Fig. 6-5B) [10]. Photoreactive derivatives of the polypeptide toxins of scorpion venom have been attached covalently to Na⁺ channels in intact cell membranes, allowing direct identification of channel components without purification. Reversible binding of saxitoxin and TTX to their common receptor has been used as a biochemical assay for the channel protein. Solubilization of excitable membranes with nonionic detergents releases the Na⁺ channel, and the solubilized channel can be purified by chromatographic techniques that separate glycoproteins by size, charge and composition of covalently attached carbohydrate. Using this general strategy, Na⁺ channels have been purified from the electric organ of the electric eel and from mammalian brain and skeletal muscle [11–13].

Covalent labeling of Na⁺ channels in intact excitable cells or membranes and purification of channels solubilized by nonionic detergents result in identification of a large glycoprotein with a molecular weight of 260,000 as the principal component. In eel electroplax, it appears to be the only protein component, but in mammalian brain this large α subunit is associated with two additional polypeptides: β₁, with a molecular weight of 36,000 and β₂, with a molecular weight of 33,000. In skeletal muscle, the α subunit is associated with only the β₁ subunit.

Figure 6-6A illustrates the most probable arrangement of the subunits of the Na⁺ channel from brain. The α subunit is a transmembrane polypeptide since it has sites for attachment of several carbohydrate chains and for binding of neurotoxins on the external surface of the channel and sites for phosphorylation by protein kinases on the intracellular surface. Since this single polypeptide suffices to form a channel by itself (see below), a transmembrane orientation is essential to its function. The β₁ and β₂ subunits also are glycosylated heavily. The β₁ subunit is attached noncovalently to the α subunit, while the β₂ subunit is attached covalently via a disulfide bond. The β subunits are integral membrane glycoproteins that interact with the phospholipid bilayer. Much of the carbohydrate on the channel subunits is sialic acid, which contributes to their strong net negative charge. Glycosylation is required for normal biosynthesis and assembly of the functional channel in neurons. If glycosylation is inhibited, newly synthesized α subunits are degraded rapidly and are not inserted into the cell surface membrane.

Purified Na⁺ channels are functional after reconstitution

An important step in the study of a purified membrane protein is to reconstitute its function in the pure state. This has been accomplished in two ways for the Na⁺ channel. In the first approach, purified channels were incorporated into vesicles of pure phospholipid. Activation of the reconstituted channels by treatment with the neurotoxin veratridine markedly increased the permeability of the vesicles to Na⁺. The purified channels retain the ion selectivity and pharmacological properties of native channels. In the second approach, ion conductance mediated by single purified channels was measured electrically. Channels reconstituted in phospholipid vesicles were studied directly with patch-clamp methods or incorporated into planar phospholipid bilayer membranes by fusion. The individual purified channels retained the single-channel conductance, ion selectivity and voltage dependence of activation and inactivation that are characteristic of native channels. Hence, purified Na⁺ channels seem to contain all of the functional components necessary for electrical excitability.

Primary structures of Na⁺- channel subunits have been determined using cDNA cloning

The amino acid sequences of the Na⁺ channel α, β₁ and β₂ subunits have been determined by cloning DNA complementary to the mRNA encoding them using antibodies and oligonucleotides developed from work on purified Na⁺ channels. The amino acid sequence of the subunits is then deduced from the nucleotide sequence of the mRNA encoding them [14,15]. The primary structures of these subunits are illustrated as topological models in Figure 6-6B. The large α subunits are composed of 1,800 to 2,000 amino acids and contain four repeated domains having greater than 50% internal sequence identity. This sequence similarity implies similar secondary and tertiary structures for the four domains. Each domain contains six segments that are predicted to form transmembrane α helices and additional hydrophobic sequences that are thought to be membrane-associated and to contribute to formation of the outer mouth of the transmembrane pore (see below). In contrast, the smaller β₁ and β₂ subunits of Na⁺ channels consist of a large extracellular N-terminal segment, a single transmembrane segment and a short intracellular segment (Fig. 6-6B).

Ca²⁺ channels have a similar structure to Na⁺ channels

The general experimental strategy used in studies of the Na⁺ channel also has been applied to voltage-gated Ca²⁺ channels. Drugs and neurotoxins that act on Ca²⁺ channels have been used to identify and purify their protein components [16], and experiments to restore their function in purified form and to determine their primary structures have been completed (Fig. 6-7A) [17,18]. Ca²⁺ channels have a principal subunit, α₁, which is structurally analogous to Na⁺ channel α subunits. Ca²⁺ channels in neurons have associated α₂ and δ subunits, which form a disulfide-linked transmembrane glycoprotein complex, and β subunits, which are intracellular (Fig. 6-7A). In addition, Ca²⁺ channels in skeletal muscle have a transmembrane γ subunit. The auxiliary subunits of Ca²⁺ channels are not related in primary structure to the Na⁺ channel β subunits. Co-expression with the auxiliary α₂, β, γ and δ subunits modulates the properties of the expressed Ca²⁺ channel and can greatly increase its expression. Since voltage-gated ion channels are likely to have evolved from common ancestor proteins, comparison of the conserved structural and functional features among the principal subunits of many different channels will sharpen our view of the molecular basis of their function, as illustrated by the comparison of the mechanisms of ion permeation and gating outlined below.

K⁺ channels have been identified by genetic means

Genes that harbor mutations causing an easily detectable altered phenotype in the fruit fly Drosophila can be cloned directly from genomic DNA without information about the protein they encode. The Shaker mutation in Drosophila causes flies to shake under either anesthesia and is accompanied by loss of a specific K⁺ current in the nerve and muscle of mutant flies. By cloning successive pieces of genomic DNA from the region of the chromosome that specifies this mutation, DNA clones that encoded a protein related in amino acid sequence to the α subunit of Na⁺ channels were isolated [19]. The K⁺ channel protein is analogous to one of the homologous domains of Na⁺ or Ca²⁺ channels [19] and is thought to function as a tetramer of four separate subunits in analogy to the structure of Na⁺ and Ca²⁺ channels (Fig. 6-7B). K⁺ channels may be the ancestral voltage-gated ion channels from which the larger Na⁺ and Ca²⁺ channels evolved by two cycles of gene duplication [5]. Like the Na⁺ and Ca²⁺ channels, K⁺ channels have an auxiliary β subunit, which is an intracellular protein distantly related to Ca²⁺ channel β subunits [20].

How do the primary structures of the ion channel subunits carry out their functions?

Cloning of the cDNA encoding the Na⁺ channel subunits permits detailed tests of the functional properties of the polypeptides. cDNA clones can be used to synthesize mRNA encoding the subunits or to isolate the natural mRNA by specific hybridization. When injected into appropriate recipient cells, such as frog oocytes, isolated mRNAs can be translated to yield functional proteins. In such experiments, mRNA encoding only the α subunit of the channel is capable of directing the synthesis of functional channels in oocytes [21,22]. The α subunits, therefore, seem sufficient to carry out the basic functions of the channel. However, coexpression of the β₁ and β₂ subunits accelerates inactivation and shifts voltage dependence toward more negative membrane potentials, conferring more physiologically correct functional properties on the expressed channel [15]. Similar experiments with Ca²⁺ and K⁺ channels also show that only the principal subunits are required for channel function but that the auxiliary subunits modulate channel function [16–20]. These results indicate that the principal subunits of the voltage-gated ion channels are functionally autonomous but that the auxiliary subunits improve expression and modulate physiological properties.

How does the structure of the α subunit of the Na⁺ channel allow it to mediate selective ion transport and voltage-dependent gating? The answer remains unknown, but working hypotheses have been developed from extensive structure—function studies that help to guide current research on this problem. Both the gap junction channel and the nicotinic acetylcholine receptor (see Chap. 10) are high-conductance ion channels that form a transmembrane pore at the center of a pseudosymmetrical array of subunits. By analogy, it is believed that the transmembrane pore of the Na⁺ channel may be formed at the center of a square array of its four homologous domains (Fig. 6-6A). Formation of a transmembrane pore in the center of a symmetrical or pseudosymmetrical array of homologous structural units may be a common theme in the structure of high-conductance ion channels.

Which amino acid sequences are involved in forming the pore? For Na⁺ channels, insight into this question has come from studies of the amino acid residues required for binding of TTX, which is thought to block the outer mouth of the transmembrane pore (Fig. 6-5B). Site-directed mutagenesis experiments show that pairs of amino acid residues required for high-affinity TTX binding are located in analogous positions in all four domains near the carboxyl ends of the short hydrophobic segments between transmembrane α helices S5 and S6 [23] (Fig. 6-6B). Six of these eight residues are negatively charged and may interact with permeant ions as they approach and move through the channel. In agreement with this idea, mutation of the only two of these residues that are not negatively charged (see domains III and IV, Fig. 6-6B) to glutamic acid residues, as present in the analogous positions in the Ca²⁺ channel, confers Ca²⁺ selectivity on the Na⁺ channel [24]. Parallel results also implicated these same regions of the K⁺ channel in determining ion selectivity and conductance [25]. Evidently, these membrane-associated segments form the outer mouth and at least part of the walls of the transmembrane pore of the voltage-gated ion channels.

Structural models for voltage-dependent gating of ion channels must identify the voltage sensors, or gating charges (Fig. 6-4A), within the channel structure and suggest a plausible mechanism for transmembrane movement of gating charge and its coupling to the opening of a transmembrane pore. The S4 segments of the homologous domains have been proposed as voltage sensors [13,14,26]. These segments, which are conserved among Na⁺, Ca²⁺ and K⁺ channels, consist of repeated triplets of two hydrophobic amino acids followed by a positively charged residue. In the α-helical configuration, these segments would form a spiral staircase of positive charge across the membrane, a structure that is well suited for transmembrane movement of gating charge (Fig. 6-5). Each positive charge is proposed to be neutralized by a negative charge in one of the surrounding transmembrane segments to form a spiral array of ion pairs (Fig. 6-5). Direct evidence in favor of designating the S4 segments as voltage sensors comes from mutagenesis studies of Na⁺ and K⁺ channels [19,27]. Neutralization of positive charges results in progressive reduction of the steepness of voltage-dependent gating and of the apparent gating charge, as expected if the S4 segments are indeed the voltage sensors. At the resting membrane potential, the force of the electrical field would pull the positive charges inward. Depolarization would abolish this force and allow an outward movement of the S4 helix. A simple spiral movement, as suggested in a sliding helix model of gating [13], would have the net effect of transferring these gating charges across the membrane. Direct evidence for an outward movement of the gating charges in the S4 segments has come from experiments in which cysteine residues substituted at these positions by site-directed mutagenesis were shown to become available for chemical reaction outside the cell upon depolarization [28]. This movement of the S4 helix is proposed to initiate a more general conformational change in each domain. After conformational changes have occurred in all four domains, the transmembrane pore can open and conduct ions (Fig. 6-6C).

Shortly after opening, many voltage-gated ion channels inactivate. The inactivation process of Na⁺ channels can be prevented by treatment of the intracellular surface of the channel with proteolytic enzymes [2] or antibodies against the intracellular segment connecting domains III and IV (h in Fig. 6-6B) [29], and expression of the Na⁺ channel as two pieces with a cut between domains III and IV greatly slows inactivation [27]. A single cluster of three hydrophobic residues in this intracellular loop is required for fast inactivation [30], and inactivation is eliminated if these three hydrophobic residues are mutated to hydrophilic ones. The phenylalanine at position 1,489 is the critical residue; mutation of this single amino acid to a hydrophilic residue nearly completely blocks fast inactivation of the channel. The segment of the Na⁺ channel between domains III and IV is therefore proposed to serve as the inactivation gate by forming a hinged lid, which folds over the intracellular mouth of the pore after activation (Fig. 6-8A). The cluster of hydrophobic residues including phenylalanine 1,489 is thought to enter the intracellular mouth of the pore and to bind there as a latch to keep the channel inactivated.

A detailed model of K⁺ channel inactivation has been derived from mutagenesis experiments [31,32]. The N-terminus of the K⁺ channel serves as an inactivation particle, and both charged and hydrophobic residues are involved. A ball-and-chain mechanism [9,31,32] has been proposed in which the N-terminal segment serves as a loosely tethered ball and inactivates the channel by diffusion and binding to the intracellular mouth of the pore (Fig. 6-8B). Consistent with this mechanism, synthetic peptides with amino acid sequences corresponding to that of the inactivation particle region can restore inactivation to channel mutants in which the N-terminus has been removed. The mechanisms of inactivation of Na⁺ and K⁺ channels are similar in that hydrophobic amino acid residues seem to mediate binding of an inactivation particle to the intracellular mouth of the transmembrane pore of the activated channel in each. They differ in that charged amino acid residues are important in the inactivation particle of K⁺ channels but not Na⁺ channels and also in that the inactivation particle is located over 200 residues from the membrane at the N-terminus of the K⁺ channel compared to only 12 residues from the membrane between domains III and IV of the Na⁺ channel. It is likely that the hinged-lid mechanism of Na⁺-channel inactivation evolved from the ball-and-chain mechanism of K⁺ channels.

These models for formation of a voltage-gated transmembrane pore by the voltage-gated ion channel proteins remain speculative at this time. However, they illustrate a general scientific method: new data, such as the amino acid sequences of the ion channel subunits, inevitably lead to the formulation of specific hypotheses, which then spawn a new generation of experiments designed to test their merit. The next phase of research on the molecular properties of ion channels should give us clearer insight into the molecular basis for two of the critical functions of ion channels: selective ion conductance and voltage-dependent gating. However, our understanding will be incomplete until we know the three-dimensional structure of an ion channel at high resolution from x-ray crystallography. We can expect to see the fruits of ion-channel crystallography in the immediate future.

K⁺ channels have many relatives

K⁺ channels have many different roles in cells. For example, in neurons they terminate the action potential by repolarizing cells, set the resting membrane potential by dominating the resting membrane conductance, determine the length and frequency of bursts of action potentials and respond to neurotransmitters by opening or closing and causing prolonged changes in membrane potential [5]. These channels are regulated by a combination of voltage, G proteins and intracellular second messengers. Remarkably, evolution has created a diverse array of ion channels based on variations of the structure of voltage-gated K⁺ channels (K_v) to serve this broad range of functions (Fig. 6-7B) [33,34]. All contain the pore structure of the voltage-gated K⁺ channel attached to different regulatory domains: a G protein-binding domain in the inward rectifying K⁺ channels (K_ir), a Ca²⁺-binding domain in the Ca²⁺-activated K⁺ channels (K_Ca), a cyclic nucleotide-binding domain in the cyclic nucleotide-gated channels or a specialized gating domain in the ether-a-go-go-related K⁺ channels (K_erg). Moreover, channels of presently unknown function which have the structure of fused K_ir and K_v channels have been detected in yeast [35]. Evidently, the K⁺ channel structure presents a flexible building block of many ion channels which are specialized for specific functions in cells as different as yeast and vertebrate myocytes and neurons.

There are many other kinds of ion channels

The channels used in the action potential contrast with those generating slow potentials at synapses and sensory receptors by having strongly voltage-dependent gating. The other channels have gates controlled by chemical transmitters, intracellular messengers or by energy sources, such as mechanical deformations in touch and hearing (Box 47-1). In general, less is known about these channels than about Na⁺, Ca²⁺ and K⁺ channels of action potentials, with the exception of the nicotinic acetylcholine receptor channel of the neuromuscular junction (see Chap. 11). The ionic selectivity of these channels includes a very broad, monovalent anion permeability at inhibitory synapses; a cation permeability that is about equal for Na⁺ and K⁺ at excitatory synapses, at the neuromuscular junction and at many sensory transducers; and other more selective Ca²⁺ and Na⁺ permeabilities in other synapses. The acetylcholine receptors of the neuromuscular junction and brain, the excitatory glutamate receptors and the inhibitory GABA and glycine receptors have been solubilized and chemically purified and the amino acid sequences of their subunits determined by methods of molecular genetics (see Chap. 10). The structural features responsible for the function of these ligand-gated channels are being elucidated rapidly.

As this research shows, there is a great diversity of ion channels playing many roles in cells throughout the body. We can speculate that hundreds of genes code for structural components of channels. Beyond their functions in the nervous system, channel activity in endocrine cells regulates the episodes of secretion of insulin from the pancreas and epinephrine from the adrenal gland. Channels initiate and regulate muscle contraction and cell motility. Channels form part of the regulated pathway for the ion movements underlying absorption and secretion of electrolytes by epithelia. Channels also participate in cellular signaling pathways in many other electrically inexcitable cells. Thus, while they are especially prominent in the function of the nervous system, ion channels are actually a basic component of all animal cells, indeed of all eukaryotic cells [5].