34 IEEE TRANSACTIONS ON NANOBIOSCIENCE, VOL. 4, NO. 1, MARCH 2005 Voltage-Gated Ion Channels Francisco Bezanilla

Abstract—Voltage-dependent ion channels are membrane pro- teins that conduct ions at high rates regulated by the voltage across the membrane. They play a fundamental role in the generation and propagation of the nerve impulse and in cell homeostasis. The voltage sensor is a region of the protein bearing charged amino acids that relocate upon changes in the membrane electric field. The movement of the sensor initiates a conformational change in the gate of the conducting pathway thus controlling the flow of ions. Major advances in molecular biology, spectroscopy, and structural techniques are delineating the main features and possible struc- tural changes that account for the function of voltage-dependent channels. Index Terms—Gating currents, ionic currents, S4 segment, voltage sensor.

I. INTRODUCTION HE BIT OF information in nerves is the action potential, a T fast electrical transient in the transmembrane voltage that propagates along the nerve fiber. In the resting state, the mem- brane potential of the nerve fiber is about 60 mV (negative inside with respect to the extracellular solution). When the ac- tion potential is initiated the membrane potential becomes less negative and even reverses sign (overshoot) within 1 ms and then goes back to the resting value in about 2 ms, frequently after be- coming even more negative than the resting potential. In a land- Fig. 1. General architecture of voltage-gated channels. Top part shows the mark series of papers, Hodgkin and Huxley studied the ionic basic subunit (or domain in the case of Na and Ca channels). The gray events underlying the action potential and were able to describe background represents the lipid bilayers. The cylinders are transmembrane the conductances and currents quantitatively with their classical segments. The region between S5 and S6 forms the pore. Segments 1 through 4 are called the voltage sensor part of the channel. The C or signs in white equations [1]. The generation of the rising phase of the action indicate charges that have been implicated in voltage sensing. In the bottom potential was explained by a conductance to Na ions that in- part, a schematic view of the channel from the outside showing the assembled creases as the membrane potential is made more positive. This four subunits or domains. is because, as the driving force for the permeating ions (Na) was dipoles that respond to the membrane electric field. This was an in the inward direction, more Na ions come into the nerve and insightful prediction of ion channels and gating currents. makes the membrane more positive initiating a positive feed- Many years of electrical characterization, effects of toxins back that depolarizes the membrane even more. This positive on the conductances, molecular biological techniques, and feedback gets interrupted by the delayed opening of another improvement of recording techniques led to the identification voltage dependent conductance that is K-selective. The driving of separate conducting entities for Na and K conductances. force for K ions is in the opposite direction of Na ions; thus, These conductances were finally traced to single-membrane K outward flow repolarizes the membrane to its initial value. proteins, called ion channels, that can gate open and closed an The identification and characterization of the voltage-dependent ion conducting pathway in response to changes in membrane Na and K conductances was one of the major contributions potential (see introductory paper by P. Jordan in this issue). of Hodgkin and Huxley. In their final paper of the series, they even proposed that the conductance was the result of increased II. VOLTAGE-DEPENDENT ION CHANNELS ARE permeability in discrete areas under the control of charges or MEMBRANE PROTEINS The first voltage-dependent ion channel that was isolated and Manuscript received November 4, 2004; revised November 15, 2004. This purified was extracted from the eel electroplax where there is a work was supported in part by the U.S. Public Health Service National Institutes of Health under Grant GM30376. large concentration of Na channels [2]. Several years later, the The author is with the Departments of Physiology and Anesthesiology, D. sequence of the eel Na channel was deduced from its mRNA Geffen School of Medicine and the Biomedical Engineering Interdepartmental [3]. The first K channel sequence was deduced from the Shaker Program, University of California, Los Angeles, CA 90095 USA (e-mail: [email protected]). mutant of Drosophila melanogaster [4]. These initial sequences Digital Object Identifier 10.1109/TNB.2004.842463 were the basis to subsequent cloning of a large number of Na ,

1536-1241/$20.00 © 2005 IEEE BEZANILLA: VOLTAGE-GATED ION CHANNELS 35

Fig. 2. Comparison between a field-effect transistor (FET) and a voltage-gated ion channel. The FET transistor is represented as a p-channel device to make a closer analogy to a cation selective voltage-gated channel. (a) and (c) are the closed states; (b) and (d) are the open states. Notice that, in contrast with the FET, the gate in the voltage-gated channel indicates the actual point of flow interruption. BLM is the lipid bilayer. In the FET, D is the drain, S is the source, and G is the gate. For details, see text. K and Ca channels in many different species. Hydropathy Fig. 1). If we take a typical V-dependent K channel, its voltage plots were helpful in deciding which parts of the sequence was sensor corresponds to the gate of a p-channel FET transistor, transmembrane or intra or extracellular. A basic pattern emerged the conducting pathway of the ion channel corresponds to the from all these sequences: the functional channels are made up p-channel (Fig. 2) and the gate of the channel is the space charge of four subunits (K channels) or one protein with four homol- in the p-channel. As we will see below, this analogy is useful to ogous domains (Na and Ca channels). Each one of the do- discuss the parts of the channel but it cannot be pushed very far mains or subunits has six transmembrane segments and a pore because, although the functions are similar, the actual mecha- loop. (see Fig. 1). The fifth and sixth transmembrane segments nisms are quite different. (S5 and S6) and the pore loop were found to be responsible for ion conduction. The fourth transmembrane segment (S4) con- A. The Conducting Pathway tains several basic residues, arginines or lysines and was ini- Living cells, and in particular nerve fibers, are surrounded by tially postulated to be the voltage sensor [3] In addition S2 and a thin membrane made of a bimolecular layer of lipids. The per- S3 contain acidic residues such as aspartate and glutamate. Most meability of ions through the lipid bilayer is extremely low be- of the channels have additional subunits that modify the basic cause it takes a large amount of energy to put a charged ion function but they are not necessary for voltage sensing and ion species inside the low dielectric constant lipid bilayer [5]. The conduction. conducting pathway of ion channels lowers that energy barrier by providing a favorable local environment, thus allowing large III. THE PARTS OF THE VOLTAGE DEPENDENT CHANNEL flows under an appropriate driving force. Details of the ion con- We think of voltage-dependent channels as made of three duction pore structure, conductance, and selectivity are covered basic parts: the voltage sensor, the pore or conducting pathway in other papers in this issue [6]. What is important to emphasize and the gate. These three parts can be roughly mapped in the here is that the ion flow is proportional to the driving force for putative secondary structure (Fig. 1). The pore and the gate are the selected ion. The driving force corresponds to the difference in the S5-loop-S6 region and the voltage sensor in the S1–S4 between the voltage applied and the voltage at which there is region. As the conduction is dependent on the voltage across no flow, or reversal potential . If the channel is perfectly selec- the membrane, a useful analogy is a field effect transistor (see tive to one type of ion, say, K , then is the Nernst potential; 36 IEEE TRANSACTIONS ON NANOBIOSCIENCE, VOL. 4, NO. 1, MARCH 2005

Fig. 3. Time course of single-channel and macroscopic ionic currents. The applied voltage is in the top trace and the current recorded through one channel is shown for four different trials. The mean current is the result of thousands of trials. (a) Small depolarization to 30 mV opens the channel infrequently. (b) A large depolarization (to C30 mV) opens the channel most of the time. c is the closed state and o is the open state. otherwise, is predicted by the Goldman–Hodgkin and Katz the open times and decreasing the closed times, as seen in equation that considers concentrations and relative permeabili- Fig. 3(b). Notice also that the time elapsed between the start of ties. Knowing the conductance of the conducting pathway ,we the pulse and the first opening (first latency) is decreased for can compute the current flow through the open conducting pore the larger depolarization Apart from increasing the open times, as the magnitude of the current through the pore was increased by the larger depolarization. This is because the applied (1) is now further away from , increasing the driving force for The – curve of an open channel may be nonlinear because ion movement. Thus, this increase in current is not a result of in general is voltage dependent. increasing but is just a passive property of the open pore. An average of several thousands of repetitions gives us the macro- B. The Gate scopic ionic currents. Provided the channels do not interact, the average of thousands of repetitions is the same as having The ion conduction through the pore may be interrupted by thousands of channels operating simultaneously. The bottom closing a gate (see Fig. 2). Thermal fluctuations will close and trace (labeled mean) in Fig. 3(a) and (b) shows the macroscopic open the gate randomly and the current will have interruptions. currents for and 30 mV, respectively. Notice that the In voltage-dependent channels, the probability that the gate is turn-on kinetics is faster for a more positive potential and that open depends on the membrane potential. In the majority the current magnitude is also increased. The kinetics change is of -dependent Na ,K , and Ca channels from nerve and the result of an increased while the magnitude increase is muscle, the increases with membrane depolarization (i.e., the result of both increased and driving force. We can now decrease in the resting potential). There are a few cases, such as write the expression for the macroscopic current as Kat1 channel, where increases on hyperpolarization. The operation of the gate can be seen by recording the current (2) flowing through a single ion channel. This is possible with the patch clamp technique [7], which records currents from a very where is the channel density and is the voltage- and small patch of membrane with a small glass pipette and a low- time-dependent open probability. noise system that can resolve currents in the order of 1 pA. An example of the operation of one K channel is shown C. The Voltage Sensor in a simulation in Fig. 3. As the internal concentration of K How does become voltage dependent? It is clear that to de- is more than 10 times higher in the cell as compared to the tect changes in membrane potential, a voltage sensor is needed. extracellular space, the reversal potential for K channels is The electric field in the bilayer could be detected by electric around 80 mV. Starting with a negative membrane potential or magnetic charges or dipoles that change their position ac- 100 mV , the channel is closed most of the time. A depolar- cording to changes in the field. As there is no evidence of mag- izing voltage pulse to 30 mV increases the open probability netic charges, electric charges or dipoles remain as the prime and the channel spends some time in the open state [Fig. 3(a)]. candidates. We will see below that the actual charges involved As we are dealing with one molecule, thermal fluctuations will in voltage sensing have been identified and a schematic repre- generate different responses for each repetition of the same sentation of their relocation is shown in Fig. 2(b). In the resting pulse (four of such are shown in the figure). A larger depolar- (hyperpolarized) condition, the membrane is negative inside and ization 30 mV increases the even more by increasing the positive charges are located in contact with the interior of BEZANILLA: VOLTAGE-GATED ION CHANNELS 37 the cell. Upon depolarization, the positive charges are driven toward the outside. This movement in the electric field has two consequences: it is coupled to the gate resulting in pore opening [Fig. 2(b)] and the charge translocation produces another mem- brane current that is transient in nature, called gating current.It is called gating current because it ultimately gates the channel open and closed, and it is transient because the charge locations are bound to limiting positions as they are tethered to the protein.

IV. GATING CHARGE AND THE VOLTAGE SENSOR An understanding of the voltage sensor requires a character- ization of the gating charge movement and a correlation of that movement to structural changes in the protein. In this section, we will address two functional questions. The first question is what are the kinetics and steady state properties of the gating charge movement and how does this charge movement relate to channel activation. The second is how many elementary charges move in one channel to fully activate the conductance and how does this movement occur in one channel.

A. The Gating Currents and the Channel Open Probability The movement of charge or dipole reorientation is the basic mechanism of the voltage sensor and was predicted by Hodgkin and Huxley [1]. Gating currents are transient and they only occur in the potential range where the sensor responds to the electric field; therefore, they behave like a nonlinear capacitance. In addition, as gating currents are small, to record them it is necessary to decrease or eliminate the ionic currents through the pore and eliminate the normal capacitive current required to charge or discharge the membrane. This is normally accomplished by applying a pulse in the voltage range that Fig. 4. Ionic and gating currents in Shaker-IR K channel. (a) Top traces: time course of ionic currents for pulses to the indicated potentials starting and activates the current and then subtracting the current elicited returning to 90 mV. Bottom traces: time course of the gating currents for the by another pulse or pulses in the voltage range that does not pulses indicated. Notice the difference in the amplitude calibration for ionic as activate the currents to eliminate the linear components [8], [9]. compared to gating currents. (b) The voltage dependence of the open probability € @† A and the charge moved per channel @† A. For details, see text. Using these subtraction techniques, the kinetics of Na gating currents were studied in detail in squid giant axon and other preparations where a high channel density was found. The same set of pulses from Shaker-IR with a mutation that changes combination of gating currents, macroscopic ionic currents and a tryptophan into a phenylalanine in position 434 (W434F) that single-channel recordings were used to propose detailed kinetic renders the pore nonconducting [11]. Several features that are models of channel operation (see [10]). When voltage-de- characteristic of most voltage-dependent channels can be ob- pendent channels were cloned and expressed in oocytes or served. First, the ionic currents do not show significant activa- cell lines, it was possible to achieve large channel densities tion for potential more negative than 40 mV while the gating on the surface membrane and study those channels in virtual currents are visible for all the pulses applied, implying that there absence of currents from other channels. In comparison to the is charge displacement in a region of potentials where most of currents in natural tissues such as the squid giant axon, the the channels are closed. Second, the time course of activation expression systems gating currents were much larger and made of the ionic current is similar to the time course of decay of the recording easier and cleaner. In addition, the study of the the gating current. Third, the time course of the return of the pore region gave us the possibility of mutating the channel charge (gating current “tail”) changes its kinetics drastically protein to eliminate ionic conduction but maintaining the op- when returning from a pulse more positive than 40 mV, which eration of the gating currents [11]. We can illustrate the basic is precisely the potential at which ionic currents become clearly features of gating currents and their relation to ionic currents visible. The gating tails are superimposable for potentials more in recordings from Shaker K channels with fast inactivation positive than 20 mV, showing that most of the charge has removed (Shaker-IR) as shown in Fig. 4(a). moved at 20 mV. The total charge moved at each potential In this figure, two separate experiments are shown. The top may be computed as the time integral of the gating current for traces are the time course of the ionic currents recorded during each pulse. As we will see below, it is possible to estimate the pulses that range from 120 to 0 mV, starting and returning to total charge moved per channel molecule; therefore, the voltage 90 mV. The bottom traces are gating currents recorded for the dependence of the charge moved can be plotted as shown in 38 IEEE TRANSACTIONS ON NANOBIOSCIENCE, VOL. 4, NO. 1, MARCH 2005

Fig. 4(b). Knowing the number of channels present (see below), using (2) it is possible to estimate the voltage dependence of , as shown in the same figure. Fig. 4(b) establishes the rela- tionship between the charge movement and channel opening and clearly shows that the opening of the channel is not superimpos- able with charge movement as expected from a simple two-state model. A striking feature of these plots is that the relation is displaced to the left of the curve so that there is quite a large charge movement in a region where the is essentially zero. This is an expected feature of a channel that requires sev- eral processes to occur to go from closed to open, such as the classical Hodgkin and Huxley model where four independent particles are needed to be simultaneously in the active position for the channel to be open. Current recordings of the type shown in Fig. 4(a) can be used to formulate kinetic models of channel gating. These models are normally written as a collection of closed and open states in- terconnected with rate constants that, in general, are a function of the membrane potential. Initially, kinetic models were devel- oped from the macroscopic ionic currents only [1]. The addition of single-channel recordings and gating currents recordings im- poses several constraints in the possible models and in the pa- rameters fitted, thus obtaining a more robust representation of the kinetic parameters that characterize the channel. The common assumption in kinetic models is that the move- ment of the charge or dipole in the field has a finite number of low-energy positions and energy barriers in between them. According to kinetic theory, the transition rates across the en- ergy barrier are exponentially related to the negative of the free Fig. 5. Early component of the gating current. (a) Gating current recorded at 10-kHz bandwidth. (b) Gating current recorded at 200-kHz bandwidth. Notice energy amplitude of the barrier. This free energy contains non- the differences in the amplitude and time scales. (c) The time course of the electrical terms, and an electrical term such that the membrane charging of the membrane capacitance for the experiment in (b). Recordings voltage can either increase or decrease the total energy bar- done in collaboration with Dr. E. Stefani. rier, resulting in changes in the forward and backward rates of crossing the barriers. There are multiple examples in the lit- jump across the first energy barrier. Using this approach, Sigg erature with varied levels of complexity that describe well the et al. [16] computed the viscosity encountered by the gating kinetic and steady state properties of several types of voltage-de- charge in its well of energy. pendent channels (see, for example, [10], [12]–[14]). What have we learned with kinetic modeling? In fact, a great A more general approach to modeling is based on a represen- deal. It is now clear that voltage-dependent channels have mul- tation of a landscape of energy using the charge moved as the tiple closed states and, in some cases, several open states. reaction coordinate. In this type of modeling, the above-men- In general, the opening of the gate requires all four subunits tioned discrete kinetic models are also represented when the to be activated, although there are cases where intermediate landscape of energy has surges that exceed 4 kT [15]. states have been observed. Each subunit undergoes several When gating currents are recorded with high bandwidth, new transitions before reaching the active state, and in the case of components are observed [16]. Fig. 5(a) shows a typical gating K channels, they do not seem to interact until the final step current trace recorded with a bandwidth of 10 kHz. Notice that that opens the channel [17], [18]. However, in the muscle Na before the initial plateau and decay of the gating current, there channel site-directed fluorescence studies show that the inter- is a brief surge of current. When the bandwidth is increased to domain interactions are manifested prior to channel opening 200 kHz, the first surge of current is the predominant amplitude [19]. Kinetic modeling has given us a picture of the channel in as shown in the fast timescale recording of Fig. 5(b). The large terms of channel physical states with transitions between them spike of current is followed by a long plateau of current that that are regulated by voltage. Kinetic modeling is a critical step corresponds to the dip observed at 10 kHz in Fig. 5(a). However, in developing a physical model of channel operation because the area spanned by the large spike is a very small fraction all the predicted features of channel function should be repro- of the total charge moved during the entire time course of the duced by the structure of the protein and its voltage-induced gating current. The interpretation of this early component is conformational changes. best understood using the representation of the gating charge moving in a landscape of energy that undergoes a change in B. Gating Charge Per Channel tilt when the membrane potential is changed such that the When the gating charge moves within the electric field, we charge advances in its initial energy well before making the detect a current in the external circuit. The time integral of that BEZANILLA: VOLTAGE-GATED ION CHANNELS 39 current represents the charge moved times the fraction of the ion channels. At very negative , has a linear dependence on field it traverses; therefore, our measurement of gating charge , so that is exponential in , such that it increases by in does not represent the exact number of charges displaced be- only 2 mV cause it includes the arrangement of the electric field. We must keep this in mind when we represent the reaction coordinate of (7) the activation of the channel in the variable . A channel evolves from to traversing many closed and/or open states. This very steep voltage dependence explains why the – Activation of the channel corresponds to the opening of the pore curve of Fig. 4(b) shows no visible at potentials more neg- and, analogous to the chemical potential, we define the activa- ative than 40 mV; in fact, there is a finite value of of less tion potential as than 10 at potentials as negative as 100 mV. Voltage-dependent channels that have several open states (3) with charge moving between them , show plots of – that may not be used to compute . One example is the Then the activation charge displacement corresponds to the maxi K channel, activated by voltage and Ca, that shows a negative gradient of the activation potential – curve with a slope that decreases as the potential is made more negative. This result is consistent with the multiple-state (4) allosteric model proposed for this channel [27]. Gating current of one channel. The previous paragraph shows The equilibrium probabilities in each physical state of the that we can estimate the total charge that moves in one channel channel can be explicitly written using the Boltzmann distri- but does not give any ideas of how that charge movement oc- bution knowing the potential of mean force for each state curs at the single-channel level. Two limiting cases can be pro- . Then, by assigning open or closed (or intermediate states) posed. In one case, the time course of the gating current is just conductances to every state, we can wite an expression for a scaled-down version of the macroscopic gating current shown that includes the voltage dependence of . The final result of in Fig. 4. The other case assumes that the charge movement oc- the derivation [20] gives a relation between , , and curs in large elementary jumps and that the macroscopic gating the charge moving between open states current is the sum of those charge shots. In principle, if the charge is moving in discrete packets, it should be possible to (5) detect those elementary events as it has been possible to detect the elementary events of pore conduction. The problem is that Note that is the – curve shown in Fig. 5. This result is those events are predicted to be very fast and extremely small general and includes cases with any number of open and closed and not detectable above background noise with present tech- states connected in any arbitrary manner. If there is no charge niques. However, if those events do exist, they should produce movement between open states , then the – curve detectable fluctuations in the gating currents recorded from a superimposes on . In addition, it is possible to estimate , relatively small number of channels. The fluctuation analysis is the total charge per channel, by taking the limiting value of done by repeating the same voltage pulse and recording a large that makes go to zero. In the typical case of a channel that number of gating current traces. From these traces, an ensemble closes at negative potentials, we obtain mean value and an ensemble variance are obtained that allows the estimation of the elementary event (Fig. 6) (6) Gating current noise analysis was first done in Na channels expressed in oocytes [28] where indeed it was possible to detect a result that was first obtained by [21] for the special case of a fluctuations that were used to estimate an elementary charge of sequential series of closed states ending in an open state. This 2.2 . A similar analysis was done in Shaker-IR K channel method has been applied to several types of voltage-dependent [29], and the elementary charge was found to be 2.4 . This channels, and the charge per channel obtained ranges between value corresponds to the maximum shot size, and if it were per 9 and 14 [22]–[24]. subunit it would account for only 9.6 or 3.6 less than the Another way to estimate the charge per channel is to mea- 13 measured for the whole channel. This means that there sure the maximum charge from the – curve and divide by is a fraction of the total charge that produces less noise and the number of channels present ( method). The number of then it may correspond to smaller shots or even a continuous channels can be estimated by noise analysis [25] or by toxin process. Fig. 6 shows the mean and variance for a large pulse binding [26]. The value of charge per channel estimated by (to 30 mV) and a smaller depolarization (to 40 mV). For the the method was 12–13 for the Shaker K channel, a small depolarization [Fig. 6(b)], gating current noise increases value that was not different from the value obtained by the lim- by the time that more than half of the charge has moved. This in- iting slope method [24]. As the limiting slope measures only the dicates that at early times the elementary event is smaller than at charge involved in opening the channel, the agreement between longer times. The gating current decreases with two exponential the two methods implies that in the case of the Shaker channel components, which are more separated in time at small depo- there is no peripheral charge. larizations. Therefore, it is possible to attribute the small shots The large value of 12 to 13 per channel explains the very to the fast component and the large shots to the slow compo- steep voltage dependence of the superfamily of voltage gated nent. This would indicate that the early transitions of the gating 40 IEEE TRANSACTIONS ON NANOBIOSCIENCE, VOL. 4, NO. 1, MARCH 2005

likely possibility. Since the first channel was cloned, it was recognized that the S4 segment with its basic residues would be the prime candidate for the voltage sensor. By introducing mutations that neutralize the charges in S4, several studies found that there were clear changes in the voltage dependence of the conductance. However, shifts in the voltage dependence of the – curve or even apparent changes in slope in the voltage range of detectable conductance do not prove that charge neutralization is decreasing the gating charge. The proof requires the measurement of the charge per channel for each one of the neutralizations. If, after neutralization of a charged residue, the charge per channel decreases, one may assume such charge is part of the gating current. Using the methods to measure total charge per channel discussed above in Shaker, two groups found that the four most extracellular positive charges in S4 and that one negative charge in the S2 segment were part of the gating charge [24], [26] (see white symbols in Fig. 1). It is interesting to note that in several instances a neutralization of one particular residue decreased the total gating charge by more than 4 . This indicates that somehow the charges interact with the electric field where they are located such that the elimination of one charge can affect the field seen by the remaining charges. If most of the gating charge is carried by the S4 segment (4 per subunit), it gives a total of 16 . Therefore, to account for the 13 for the total channel obtained from charge/channel measurements, they all must move at least 81% of the membrane electric field .

Fig. 6. Current fluctuations in gating currents. (a) Top trace is the time course B. Movement of the Charges in the Field of a pulse to C10 mV; middle trace is the variance and bottom trace is the mean computed from several hundred traces. (b) Top trace is the time course of a Knowing the residues that make up the gating charge is a big smaller pulse to 40 mV; middle trace is the variance; lower trace is the mean advance because it makes it possible to test their positions as a computed from hundreds of traces. Modified from [28]. function of the voltage and, thus, infer the possible conforma- current are made up of several steps, each carrying a small el- tional changes. ementary charge. The total movement of charge in the early As we will see below, the literature contains a large number of transitions would have to account for the 3.6 needed to make papers (for a review, see [30] and [31]) with many types of bio- the total of 13 per channel. We conclude from these experi- physical experiments testing the accessibility, movement, and ments that the gating current of a single channel is made up of intramolecular distance changes. With all these data, models small shots at early times followed big shots of currents that in of the voltage sensor movement were proposed that could ac- the average show two exponential decays. count for all the experimental observations but in total absence of solid information on the three-dimensional (3-D) structure V. S TRUCTURAL BASIS OF THE GATING CHARGES of the channel. In Section VI, we will discuss many of the bio- In this section, we will address the relation between the func- physical measurements of the voltage sensor and the models that tion of the voltage sensor and its structural basis. We will first have emerged from them and the recently solved crystal struc- ask where are the charges or dipoles in the structure of the ture of KvAP [32]. channel and then how those charges or dipoles move in response to changes in membrane potential. VI. STRUCTURAL BASIS OF THE VOLTAGE SENSOR A. The Structures Responsible for the Gating A 3-D structure of the channel, even in only one conforma- Charge Movement tion, would be invaluable as a guideline in locating the charges There are many ways one could envision how the charge inside the protein and in proposing the other conformations that movement is produced by the channel protein. The putative account for the charge movement consistent with all the bio- -helical transmembrane segments have an intrinsic dipole physical measurements. We will see below that the only crystal moment that upon tilting in the field would produce an equiv- structure of a voltage-dependent channel available is not in a alent charge movement. Also, induced dipoles of amino acids native conformation. Therefore, we are still relying on data that side chains could accomplish the same. However, 13 per can only be used to propose models that are still uncertain as channel is very large, and charged amino acids become the most they lack information on the 3-D structure of the channel. BEZANILLA: VOLTAGE-GATED ION CHANNELS 41

Fig. 7. Three models of the voltage sensor. (a) Helical screw model. (b) Transporter model. (c) Paddle model. Following the chemical convention, positively charged residues are in blue. For details, see text.

A. The Crystal Structure of KvAP Along with the crystal structure of the full KvAP channel, Jiang et al. [32] solved the crystal structure of just the S1 through Recently, the long-awaited first crystal structure of a voltage- S4 region of KvAP. This crystal showed similarities but was not dependent channel, KvAP from archea Aeropirum pernix, has identical to the S1–S4 region of the full KvAP crystal structure been published [32]. The structure was a surprise because it [34]. This second crystal was docked to the pore region of the showed the transmembrane segments in unexpected positions full crystal maintaining the S2 segment within and parallel to with respect to the inferred bilayer. For example, the N-terminal the bilayer to obtain a new structure that was proposed in the was buried in the bilayer whereas it has been known to be intra- open and closed conformations as the paddle model of channel cellular; the S1–S2 linker is also buried although it has been pre- activation. It is important to note here that the structures shown viously shown to be extracellular. The S4 segment, along with in the second paper [33] do not correspond to the original KvAP the second part of S3 (S3B) formed the paddle structure that structure. In fact the proposed structures of the model depart so was intracellular and lying parallel to the bilayer, a location that much from the crystal structure that we should treat them just would be interpreted as the closed position of the voltage sensor. like any other model of activation proposed before. However, the pore gates in the same crystal structure clearly correspond to an open state. Thus, the crystal structure is in a B. Models of Sensor Movement conformation that was never observed functionally, raising the A compact way of reviewing the large body of biophysical question whether that crystal structure of KvAP is indeed repre- information on structural changes of the voltage sensor is to sentative of the native conformation of the channel in the bilayer. describe the models that are currently proposed to explain the The authors mention that the channel is inherently floppy; there- operation of the voltage sensor. fore, in obtaining the structure, it was necessary to cocrystallize Fig. 7 shows schematically three classes of model that have it with Fab fragments attached to the S3–S4 loop of the channel. been proposed to explain the charge movement in voltage-de- This raises the possibility that the fragments may have distorted pendent channels. In all cases, only two of the four sensors are the channel structure that was in the final crystal. The authors shown and the coupling from the sensor to the gate is not shown functionally tested the structure by incorporating the channel explicitly. In addition, the charge that moves resides completely in bilayers and recording the currents through the channel after in the first four charges of the S4 segment. The common feature Fab fragments were added to the inside or to the outside of the of all three models is that the charge is translocated from the channel. The Fab fragment did not attach from the inside but inside to the outside upon depolarization. However, there are only from the outside, showing that the position of the S3–S4 important differences as of how those charges relocate in the shown in the crystal structure (obtained in detergent) did not protein structure. represent a native conformation of the channel in the bilayer Fig. 7(a) shows the conventional helical screw model [35], [33]. [36]. Although there are variations on this model, the general 42 IEEE TRANSACTIONS ON NANOBIOSCIENCE, VOL. 4, NO. 1, MARCH 2005 idea is that upon depolarization the S4 segment rotates along to be protected, contrary to the location proposed in the paddle its axis and at the same time translates as a unit perpendicular model. to the membrane, thus changing the exposure of the charges Laine et al. [46] found that the extracellular part of the S4 from the intracellular to the extracellular solution, effectively segment gets within a few Ångstroms of the pore region in the translocating 4 per subunit [see Fig. 7(a)]. In the original open state of Shaker-IR channel. This result is consistent with version of the model, the change of exposure required a large the helical screw and the transporter model but is inconsistent 16- translocation of the S4 segment, and the positively charged with the paddle model presented in [33] because in their model arginines were making saline bridges with aspartate or gluta- the extracellular portion of S4 is well separated from the rest mates residues that had to be broken to initiate the movement. In of the channel protein, giving a distance of about 98 be- more recent versions [37]–[39], the charges are in water crevices tween segments across the pore region. In experiments using in both the closed and open position, decreasing the amount of resonance energy transfer with lanthanides [47], [48] or with translation required of the S4 segment. organic fluorophores [49], that distance was found to be around Fig. 7(b) shows the transporter model [40]–[42]. In the closed 50 A, also in agreement with measurements done with tethered position, the charges are in a water crevice connected to the TEA derivatives [50] These results indicate that the S4 segment intracellular solution, and in the open position they are in an- is not extended into the bilayer but is against the bulk of the other water crevice connected to the extracellular solution. The channel protein. If the paddle model were modified, so that the translocation of the charges is achieved by a tilt and rotation of S4 segments would be almost perpendicular to the plane of the the S4 segment with little or no translation. In this case, the field bilayer in the open state and almost parallel to the bilayer in the is concentrated in a very small region that changes from around closed state but still flush against the rest of the protein, it would the first charge in the closed state to the fourth charge in the open be consistent with the distance constraints just mentioned. In a state. more recent report, the MacKinnon group made such modifica- Fig. 7(c) shows the paddle model introduced by the tion at least for the open-inactivated state [51]. However, even MacKinnon group [33] where the S4 segment is located with these modifications the paddle model locates the charges in in the periphery of the channel and the charges are embedded the bilayer, a proposal that is inconsistent with the EPR results in the bilayer. The S4 segment makes a large translation such [45] and several other biophysical experiments [52], [54]–[63]. that the most extracellular charge goes from exposed to the Locating the charges in the bilayer has been a subject of inten- extracellular medium in the open state to completely buried in sive debate because the energy required in moving a charge into the bilayer in the closed state. the low dielectric constant bilayer is extremely high [5]. If pos- The helical screw and transporter models are similar but they itively charged gating charges are neutralized by making salt differ dramatically from the paddle model in that the gating bridges with acidic residues, the energy decreases [5] but there charges in the paddle model are embedded in the bilayer, while would not be any gating charge movement. in the helical screw and transporter models, the charges are Evidence obtained by charge measurements in the squid axon surrounded by water, anions, or making salt bridges. In con- sodium channel has shown that the gating charges do not move trast with the transporter model, the helical screw model has in the bilayer [52] In these experiments, addition of chloroform in common with paddle models the large translation of the S4 increased the kinetics of translocation of the hydrophobic ion segment. dipycrilamine while it did not change the kinetics of the sodium In Sections VI-C–VI-E, we will review biophysical experi- gating currents. The conclusion was that the gating charge, un- ments that were designed to test the topology of the channel and like hydrophobic ions, does not move in the bilayer. Ahern and the extent of the conformational changes of the gating charge. In Horn [38] have explored this subject in more detail. As in the the absence of a native crystal structure, these experiments are paddle model the S4 segment is in the bilayer, they reasoned that the only data that we can use to support or reject the available on addition of more charges in the S4 segment, the net gating models of voltage sensor operation. charge should increase. Their results show that the charge ad- dition at several positions did not increase the gating charge, indicating that only the charges in aqueous crevices are respon- C. The Topology of the Channel and Gating Charge Location sible for gating and can sense the changing electric field. Both Alanine and tryptophan scanning have been used to infer the these experimental results are hard to reconcile with the paddle relative positions of the transmembrane segments [43], [44], model where the voltage sensor is immersed in the hydrophobic and the results indicate that the S1 segment is in the periphery core of the lipid bilayer. of the channel. This result is at odds with the recent report by The S1–S2 loop has been clearly located in the extracellular Cuello et al. [45] where using electron paramagnetic resonance region by several criteria. One is that a glycosylation site in (EPR) scanning, they found that in KvAP the S1 segment is Shaker occurs in this loop [53]. In addition, fluorescence signals not exposed to the bilayer but rather surrounded by the rest of from fluorophores attached in this loop are consistent with the protein. This difference could be an inherent limitation in this region being extracellular [54], as well as the recent EPR the ala or tryp scanning techniques that test the function of the scanning of KvAP [45]. These results are in agreement with the channel, or it could be a genuine difference between eukaryotic helical screw and transporter models but are again inconsistent and prokaryotic channels. On the other hand, their report shows with the location proposed in the paddle model. In the KvAP that the S4 segment seems to be exposed to the bilayer, in agree- crystal structure the S1–S2 linker is buried in the bilayer with ment with the paddle model although the charged residues seem an S2 segment almost parallel and also buried in the bilayer. BEZANILLA: VOLTAGE-GATED ION CHANNELS 43

In fact, this orientation of the S2 segment was the basis to that the three outermost charges change exposure with voltage dock the S1–S4 crystal to the rest of the crystal to propose and are consistent with the idea of a conformational change that the paddle model. translocates charges from inside to outside upon depolarization giving a physical basis to the gating currents. D. Voltage-Induced Exposure Changes of the S4 Segment and These studies were also done in the Shaker-IR K channel Its Gating Charges [57]–[59] and showed the same trend: the outermost charges Testing the exposure of the gating charges in the intra- or ex- were exposed to the outside upon depolarization and the inner- tracellular medium would give us an idea whether their voltage- most to the inside on hyperpolarization. While some residues induced movement takes them out of the protein core or from were inaccessible from one side to the cysteine-reacting com- the lipid bilayer. There have been three different approaches to pound on changing the voltage, there were a few that reacted test exposure. The first method, cysteine scanning mutagenesis, when the reactive agent was added to the other side. consists of replacing the residue in question by a cysteine and The conclusion from the cysteine scanning experiments is then test whether a cysteine reagent can react from the extra or that the residues bearing the gating charges change their expo- intracellular solution and whether that reaction is affected by sure with the voltage-dependent state of the channel. Another voltage [55]–[59]. The second method, histidine scanning mu- very important conclusion is that the charges are exposed to the tagenesis, consists on titrating with protons the charge in the solutions and not to the lipid bilayer, since the SH group in the residues. In this case, as the pKa of the arginine is very high, cysteine has to be ionized to react with the cysteine modifying histidine was used as a replacement allowing the titration in a reagent. This would be energetically unfavorable in the low di- pH range tolerated by the cell expressing the channel [41], [60], electric medium of the bilayer. It is important to note that the [61]. The third method consists of replacing the residue in ques- reactivity of thiol groups on some the sites tested was compa- tion with a cysteine, followed by tagging it with a biotin group rable to their reactivity in free solution. and then testing whether streptavidin can react from the inside Histidine scanning mutagenesis. Testing the titration of the or outside depending on the membrane potential [33]. The out- charged residues is a direct way to address exposure of the come of all these experiments is that indeed the charged residues charges to the solutions. In this approach, each arginine or in S4 undergo changes in exposure when the membrane poten- lysine is exchanged to a histidine residue that can be titrated tial is changed. However, it is important to look at each of these in a pH range that is tolerated by the expression system. The procedures with their limitations and compare their results be- titration of the histidine with a proton can be easily detected cause that may reveal many details of the actual movement of as a change in the gating current, provided the ionic current the charge. is blocked. For this reason, most of these experiments were Cysteine scanning mutagenesis. The experiments by Yang carried out in the nonconducting Shaker-IR K channel that and Horn [55] were the first to test the accessibility of charged bears the W434F mutation. groups to the extra and intracellular solutions and its depen- The logic of this procedure can be understood by taking some dence on membrane potential. The idea of these experiments limiting cases. is to test whether a cysteine reacting moiety can attach to an en- gineered cysteine in the channel depending where the moiety is 1) If the histidine is not exposed to the solutions and/or if and what the electric field is. This method works provided the it does not move in the field, it would not be possible to attachment induces a detectable change in the channel currents. titrate it from either side or under any membrane potential; In addition, the attachment of the moiety will depend on the therefore, no change in the gating currents are observed. local pH and state of ionization of the cysteine residue. Yang 2) The histidine can be exposed to the inside or to the out- and Horn [55] showed that the reaction rate of MTSET to the side depending on the membrane potential and thus on cysteine replaced most extracellular charge of the S4 segment the conformation of the voltage sensor. In this case, if of the fourth domain of a sodium channel depended on mem- a pH gradient is established, every translocation of the brane potential. To conclude that a particular site changes its voltage sensor would also translocate a proton, thus pro- exposure, the reaction rate must differ by more than an order ducing a proton current. This proton current would be of magnitude between the two conditions. This is important be- maximum at potentials where the sensor is making most cause if one just measures whether the reaction occurred or not, excursions which occurs around the midpoint of the – one may be sampling a rarely occurring conformational state. curve. At extreme potentials, the gating current would be Their result showed for the first time that the accessibility of affected but no steady proton current would be observed. this group changed with voltage or alternatively that the ioniza- Therefore, the – curve of such proton current would be tion state was changed by voltage. This work was expanded to bell-shaped. the deeper charges in the S4 segment of domain IV of the Na 3) One particular conformation of the sensor locates the his- channel [56]. The conclusion was that upon depolarization the tidine as a bridge between the internal and external so- most extracellular charge was exposed and that at hyperpolar- lutions forming a proton selective channel. In this case ized potential it became buried. In addition, the two following a steady current would be observed that would have an charges could be accessed from the inside at hyperpolarized po- almost linear – curve in the range of potentials where tential and from outside at depolarized potential. The next two that conformation is visited and would be zero otherwise. deeper charges were always accessible from the inside regard- The analysis of histidine scanning was done in Shaker-IR K less of the membrane potential. These results strongly suggest channels and six charges were tested starting from the most 44 IEEE TRANSACTIONS ON NANOBIOSCIENCE, VOL. 4, NO. 1, MARCH 2005

next two arginines are on the back side of the S4 helix and are not visible. The hydrophobic region, which includes other seg- ments and the bilayer, is simply shown as a gray area. In the closed state, the first four charges are in contact with the in- ternal solution by way of a water crevice (up arrow); however, the most extracellular charge (R362) is in the boundary between the intracellularly connected water crevice and the extracellular solution. If this arginine were replaced by a histidine, it would form a proton pore by making a bridge between both solutions. Upon depolarization, the S4 segment changes its tilt, and all four charges become exposed to a water crevice (down arrow) con- nected to the extracellular solution. Now the fourth charge is in the boundary between the extra and intracellular solutions so that a histidine in this position would form a proton pore. No- tice that R365 (and also R368) change exposure in such a way that if each one were replaced by a histidine in the presence of a proton gradient, every transition would be able to shuttle one proton from one side to the other. It was found that R371H is a transporter but it can also form a proton pore at large depolar- ized potentials [61]. The consequence of these results is that in the two extremes positions the electric field is concentrated [62] in a very narrow region of the protein: near 362 in the closed position and near 371 in the open position. This means that there is no need for a large movement of the voltage sensor to transport a large amount of charge. Asamoah et al. [63] have shown that indeed the field is concentrated in this region by using a cysteine reactive elec- trochromic fluorophore. By attaching this fluorophore in dif- ferent regions of the channel and comparing the fluorescence Fig. 8. Results of histidine scanning mutagenesis. (a) Each mutant is listed signal induced by voltage changes with the same electrochromic with the type of current observed and their accessibility to the internal and group in the bilayer, the field in the S4 region was found to be external solutions. In the case of K374H and R377H, the histidine may not be accessible and/or they do not move in the field. (b) Interpretation of the at least three times more intense. histidine scanning experiments with the transporter model based on a molecular The results of histidine and cysteine scanning experiments model built with KvAP and KcsA crystal structures [42], [80]. A change in suggest that the charged arginine residues are stabilized by water tilt of the S4 segment exposes the first four charges to the outside in the depolarized condition and to the inside in the hyperpolarized condition. In the and possible anions residing in the water filled crevices. It has hyperpolarized condition, there is a very narrow region bridged by histidine in been shown by Papazian and collaborators [64], [65] that the position 362 and in the depolarized condition a bridge is formed by histidine in acidic residues in S2 and S3 segments play a stabilizing role in position 371. For details, see text. the structure of Shaker-IR channel but they could also lower the energy of the ariginines in the crevices and possibly contribute extracellularly located [41], [60], [61]. Fig. 8(a) shows the re- to the gating charge by focusing the field in that region. sults where the accessibility of each of the mutant is shown. The Biotin and streptavidin scanning. Biotin has a length of about first four charges are accessible from both sides depending on 17 and its cysteine modifying derivative can react with cys- the membrane potential, while the next two are not titratable. teines engineered in the channel protein. In addition, the biotin These results are consistent with what we know about the role group binds avidin, a large soluble protein molecule that is not of each charge. Only the first four charges, which are responsible expected to cross the membrane. Jiang et al. [33] studied sev- for most of the gating charge, seem to translocate, while the next eral positions by mutating residues to cysteine in S3b and S4 two either do not move in the field or are never accessible from segments of KvAP and attached to it a biotin molecule. After the solution. The results also show that some of the charges that incorporating these channels into bilayers, they tested whether appeared buried to the cysteine scanning mutagenesis method the currents were affected by avidin in the external or internal are accessible to protons indicating that they are pointing into solutions. They found that, depending on the site of biotin at- deep crevices that are too narrow for the cysteine-modifying tachment, the currents were decreased by externally or inter- reagent but large enough for protons to reach. nally applied avidin. There were two sites, (121 and 122) lying Fig. 8(b) shows how the results could be explained in terms of in between the second and third charges of KvAP, where they an actual structure of the S4 segment with its associated mem- saw inhibition by avidin from both sides. They indicated that in brane segments and bilayer, based on a transporter-like molec- the conventional model (helical screw or transporter models) it ular model [42], [80] built with the crystal structures of KvAP would be extremely difficult to relocate the linker of the biotin and KcSA. The alpha-helical S4 segment is represented by a along with the charges moving within the protein core. There- cylinder showing the first four charged arginines, as labeled. The fore, they proposed that those sites must move a large distance BEZANILLA: VOLTAGE-GATED ION CHANNELS 45 within the bilayer to reach for the avidin present in the solu- However, several important qualitative results have been ob- tion and gave them the basic restrictions on the extent that the tained. For example, it has been found that the kinetics of the paddle must move. Although this result is consistent with the fluorescence changes in S4 are slower than around the S1–S2 paddle model, it is not inconsistent with the transporter model linker, suggesting that there might be earlier conformational because those residues may in fact be facing the bilayer [42] and changes that precede the main conformational change normally thus allowing the biotin to reach to either side of the bilayer to attributed to the S4 segment [67]. Also, the time course of attach to the avidin. It should be noted that no reaction rate was fluorescence changes near the pore region are much slower than measured in those experiments [33], so that it is also possible channel activation and their kinetics can be traced to another that their inferred conformations were rarely visited. gating process called slow inactivation in Shaker K channel [67], [71]. Probably one of the most informative results have been ob- E. Fluorescence Spectroscopy Reveals Conformational tained in the sodium channel because, as this technique detects Changes of the Voltage Sensor local changes, it has been possible to distinguish specific func- tions for each one of the four domains of the Na channel. Fast Site-directed fluorescence is a powerful technique to follow inactivation is another gate that operates by blocking the channel conformational changes in a protein. In the case of voltage de- pore [1], [72], [73]. By labeling the S4 segment of each domain pendent channels, the idea is to label specific sites of the channel of the Na channel, it was found that the gating charge immobi- protein with a fluorophore and measure changes in fluorescence lization produced by inactivation only occurred in domains III induced by changes in the field. The salient feature of this tech- and IV,thus locating the regions of the channel that interact with nique is that the changes measured reflect local changes in or the inactivating particle [73]. The kinetics of the fluorescence of near the site where the fluorophore is located in the protein as sites in S4 was found to be very fast in domains I–III but slower opposed to electrical measurements that reflect overall confor- in domain IV,indicating that domain IV followed the other three mations of the protein. Two main types of measurements have domains [75]. Finally, by comparing the effect of a perturbation been done with site-directed fluorescence. One is the detec- in one domain to the fluorescence signal in another domain, it tion of fluorescence intensity changes of one fluorophore in the was found that all four domains of the Na channel move in labeled site, and the other is the estimation of distance and dis- cooperative fashion. This result is important in explaining why tance changes between two fluorophores using fluorescence res- sodium channels are faster than potassium channels, an abso- onance energy transfer (FRET). lute requirement in eliciting an action potential [19]. For more Site-directed fluorescence changes. In these experiments, the details on the voltage-dependent sodium channel, see the paper site of interest is mutated to a cysteine and then is reacted to by French and Zamponi in this issue. a fluorophore that has a cysteine reactive group and the time Resonance energy transfer. By labeling the protein with a course of fluorescence is monitored during pulses applied to donor and an acceptor fluorophore, it is possible to estimate the open or close the channel [66], [67]. To detect a fluorescence distance between them using the Förster theory of dipole–dipole change, the conformational change must change the environ- interaction [76]. Depending on the fluorophore pair used, those ment around the fluorophore so that the intensity changes be- distances can be from a few to about 100 , thus enabling cause of spectral shifts, quenching, or dye reorientation. In most the measurements of intramolecular distances and changes in cases, the changes in fluorescence have been traced to changes distances. in the quenching environment around the fluorophore [67], [68], This technique was used to estimate distances and distance and in a few cases it is produced by spectral shifts of the fluo- changes between subunits in Shaker-IR channel by Cha et al. rophore as the hydrophobicity of the environment changes with [47] and Glauner et al. [49]. Cha et al. [47] used a variant of the change in conformation [54]. In some cases, the presence of FRET called LRET that uses a lanthanide (terbium) as a donor quenching groups in the protein is crucial in obtaining a signal. and has the advantage that the estimation of the distances are For example, in the bacterial channel NaChBac, the signals are more accurate mainly because the orientation factor is bound very small or in some sites not detectable [69], although there between tight limits giving an uncertainty of 10 [77]). are four charges in the S4 segment following the classical pat- Both studies [47], [49] showed that the distances between tern and the total gating charge was recently measured to be S4 segments were around 50 and that it did not change very about 14 [70]. This result has been traced to the lack of much from the closed to the open states. The maximum distance quenching groups in the structure of this channel [69]. To obtain change measured by Cha et al. [47] was about 3 while Glauner signals from sites near the S4 segment upon changes of mem- et al. [49] measured about 5 . These measurements were all brane potential, a requirement seems to be the presence of gluta- done between subunits; therefore, they do not completely rule mate residues in the nearby region that act as quenching groups out a translation across the bilayer, as proposed in the paddle [69]. model. Fluorescence changes in site-directed fluorescent labeling In Shaker K channel, the linker between S3 and S4 is made have provided information of local conformational changes of about 30 residues. At least six of the residues close to the around the S4 segment, the S3–S4 linker, the S1–S2 linker, and extracellular part of S4 seem to be in alpha-helical conformation the pore region as a result of changes in membrane potential [78] suggesting that S4 is extended extracellularly as an alpha [66], [67], [71]. In the absence of 3-D structure, the interpre- helix. This would explain why Cha et al. [47] measured changes tation of these fluorescence changes is not straightforward in distance in the S3–S4 linker as a result of membrane potential because the exact location of the fluorophore is unknown. changes. One of those changes is a rotation in the linker and 46 IEEE TRANSACTIONS ON NANOBIOSCIENCE, VOL. 4, NO. 1, MARCH 2005 the other is a decrease in the tilt of the S4 with its extension the bilayer, the two main features of the paddle model become upon depolarization, providing a possible mechanism for charge inconsistent with the available data. translocation (see Fig. 8). The question of how much translation the S4 undergoes VII. COUPLING OF THE SENSOR TO THE GATE across the bilayer with depolarization has been approached In contrast to a field-effect transistor where gating of the with two other variants of FRET. In the first series of experi- channel is produced by a change in space charge, the gate in ments [79], green fluorescent protein (eGFP) was inserted after voltage-gated channels seem to be a mechanical obstruction to the S6 of Shaker K channel and was used as the donor to an flow [30]. The crystal structure of the bacterial channel KcSA acceptor attached in the extracellular regions of the channel. reveals a closed state of the channel and Perozo et al. [85] found In this case, the acceptor was a sulforhodamine with an MTS that when it opens the S6 makes a scissor-like action allowing reactive group that can react with an engineered cysteine in the ions to go through. The crystal structure of MtsH, another channel but it also can be cleaved off with a reducing agent. This prokaryot channel, shows an open pore where a glycine in the allows the measurement of donor fluorescence (intracellularly S6 segment is shown to break the S6 in two segments [86]. located) in presence and absence of acceptor (in the extracel- This led MacKinnon to propose that the gate opens when the lular regions of the channel) to compute the energy transfer S6 segment is broken and the intracellular part is pulled apart and the distance. These measurements were done in multiple [87]. In the crystal structure of the voltage-dependent prokaryot sites of the S1–S2 loop, S3–S4 loop, and S4 segment under channel, KvAP the pore is in the open conformation by a break depolarized and hyperpolarized conditions giving changes in in the S6 segment in a glycine residue [32]. In the case of the distance that did not exceed 2 . In fact, some sites increased eukaryotic Shaker K channel, there is a PVP sequence in the while others decreased their distance to the intracellular donor S6 segment that has been proposed to be the actual gate [88]. upon depolarization, indicating that there is little translation of In addition to the main gate formed by the bundle crossing the S4 across the membrane. of the S6 segments, there is now very good evidence that the The second method used the hydrophobic ion dipycrilamine selectivity filter can also stops conduction, thus introducing (dpa) as an acceptor that distributes in the edges of the bilayer another gate in series [89]. However, there is no evidence or a according to the membrane potential [52], [80]. The donor was physical mechanism to couple this filter gate to the movement rhodamine attached to specific sites in the S4 segment. The ex- of the sensor. periment predicts a clear distinction for the outcome depending The question of how and what kind of physical movement of whether the S4 does or does not make a large translation across the sensor, mainly the S4 segment, couples the opening of the the membrane. If the S4 segment undergoes a large translation pore gate is far from resolved. Most of the proposals, including across the bilayer such that the donor crosses its midpoint, then the paddle model, suggest that the change in position of the S4 a transient fluorescence decrease is expected because dipycril- couples via the intracellular S4–S5 linker to change the position amine and the fluorophore start and end in opposite sides of the of the S5 segment that in turn allows the opening of S6. In the membrane but there is a period where both donor and acceptor transporter model [42], [80] the change in tilt of the S4 segment reside simultaneously in both sides of the membrane increasing carries the S5 segment away from S6, allowing the break at the transfer and consequently decreasing donor fluorescence. On glycine and thus opening the pore. the other hand, if there is no crossing of the bilayer midline by the donor, the fluorescence will increase if the donor is above VIII. CONCLUSION the midline or decrease if it is below, but no transient should be The main component of the sensor of voltage-gated channels observed. The results of at least four sites are consistent with no is the S4 segment with its basic residues moving in the field. The crossing of the midline strongly suggesting that the S4 does not movement of the sensor in response to changes in the electric make a large translation upon depolarization [80]. field produces the gating currents. The study of gating currents, Finally, one more experiment using LRET confirms the lack single-channel currents, and macroscopic currents has gener- of large translation of the S4 [81]. In this case, the donor is Tb ated detailed kinetic models of channel operation. The sensor (in chelate form) attached in several sites of the S4 segment and couples to the gate possibly via the S4–S5 linker. The channel the S3–S4 linker while the acceptor is in a toxin that blocks the fully opens after all sensors have moved, and there is little coop- pore of Shaker from the extracellular side. Results show that the erativity in the early movements of the sensor in the K channel changes in distance between the toxin and several tested sites but strong positive cooperativity in the Na channel. The results in the S3–S4 linker and S4 segment do not exceed 2 upon of a large variety of biophysical experiments have helped in de- depolarization. lineating the conformational changes of the sensor. These data A recent molecular dynamic calculation of the voltage has shown that the S4 segment does not undergo large move- sensor of a K channel under the influence of an electric field ments. Rather, the charges are in water crevices, and they only also supports a conformational change that involves minimum move a small distance because the field is focused in a narrow translation [82]. region within the protein core. In the absence of a crystal struc- In summary, FRET experiments are inconsistent with a large ture representative of the channel in its native form, we believe transmembrane displacement of the S4 segment in response to a that the data support the transporter model or some of the most voltage pulse. As many other experiments, presented above and recent versions of the helical screw model of the voltage sensor. discussed in recent papers and reviews [34], [40], [46], [83], A detailed understanding of voltage gated channels means [84], do not support the idea that the charged residues are in that we can represent the landscape of energy of the physical BEZANILLA: VOLTAGE-GATED ION CHANNELS 47

states of the channel at atomic resolution together with the struc- [16] D. Sigg, F. Bezanilla, and E. Stefani, “Fast gating in the Shaker K tural changes evoked by the electric field. To achieve this goal, it channel and the energy landscape of activation,” Proc. Nat. Acad. Sci., vol. 100, pp. 7611–7615, 2003. will not only require a static, high resolution 3-D structure, but [17] R. Horn, S. Ding, and H. J. Gruber, “Immobilizing the moving parts also a detailed description of the kinetics of the voltage-induced of voltage-gated ion channels,” J. Gen. Physiol., vol. 116, pp. 461–476, 2000. conformational changes. The latter are expected to be obtained [18] L. M. Mannuzzu and E. Y. Isacoff, “Independence and cooperativity in with spectroscopic and computational techniques. Fluorescence rearrangements of a potassium channel voltage sensor revealed by single and EPR spectroscopies have started to unravel some of the de- subunit fluorescence,” J. Gen. Physiol., vol. 115, pp. 257–268, 2000. [19] B. Chanda, O. K. Asamoah, and F. Bezanilla, “Coupling interactions be- tails of conformational changes during gating but the study of tween voltage sensors of the sodium channel as revelaed by site-specific detailed kinetics is still developing. For example, in the same measurements,” J. Gen. Physiol., vol. 123, pp. 217–230, 2004. [20] D. Sigg and F. Bezanilla, “Total charge movement per channel: The re- way that single-channel current recordings were critical in un- lation between gating displacement and the voltage sensitivity of activa- derstanding the operation of the pore and the gate, single-mole- tion,” J. Gen. Physiol., vol. 109, pp. 27–39, 1997. cule fluorescence is expected to show the local conformational [21] W. Almers, “Gating currents and charge movements in excitable mem- branes,” Rev. Physiol. Biochem. Pharmacol., vol. 82, pp. 96–190, 1978. changes during voltage sensor operation and channel gating. [22] B. Hirschberg, A. Rovner, M. Lieberman, and J. Patlak, “Transfer of These techniques are advancing very rapidly, and some results twelve charges is needed to open skeletal muscle Na channels,” J. Gen. Physiol., vol. 106, pp. 1053–1068, 1996. have already been published as just the conformational change [23] F. Noceti, P. Baldelli, X. Wei, N. Qin, L. Toro, L. Birnbaumer, and E. [90], [91] or even correlated to current recordings [92], [93]. Stefani, “Effective gating charges per channel in voltage-dependent K and Ca channel,” J. Gen. Physiol., vol. 108, pp. 143–155, 1996. [24] S.-A. Seoh, D. Sigg, D. M. Papazian, and F. Bezanilla, “Voltage-sensing ACKNOWLEDGMENT residues in the S2 and S4 segments of the Shaker K channel,” Neuron, vol. 16, pp. 1159–1167, 1996. The author would like to thank Drs. R. Blunck and B. Chanda [25] N. E. Schoppa, K. McCormack, M. A. Tanouye, and F. J. Sigworth, “The for reading the manuscript and for their helpful and insightful size of gating charge in wild-type and mutant Shaker potassium chan- nels,” Science, vol. 255, pp. 1712–1715, 1992. comments. [26] S. K. Aggarwal and R. MacKinnon, “Contribution of the S4 segment to gating charge in the Shaker K channel,” Neuron, vol. 16, pp. 1169–1177, 1996. REFERENCES [27] F. T. Horrigan, J. Cui, and R. W. Aldrich, “Allosteric voltage gating of [1] A. L. Hodgkin and A. F. Huxley, “A quantitative description of mem- potassium channels I: mSlo ionic currents in absence of Ca ,” J. Gen. brane current and its application to conduction and excitation in nerve,” Physiol., vol. 114, pp. 277–304, 1999. J. Physiol., vol. 117, pp. 500–544, 1952. [28] F. Conti and W. Stuhmer, “Quantal charge redistribution accompanying [2] W. S. Agnew, S. R. Levinson, J. S. Brabson, and M. A. Raftery, “Pu- the structural transitions of sodium channels,” Eur. Biophys. J., vol. 17, rification of the tetrodotoxin-binding component associated with the pp. 53–59, 1989. voltage-sensitive sodium channel from electrophorus electricus electro- [29] D. Sigg, E. Stefani, and F. Bezanilla, “Gating current noise produced by plax membranes,” Proc. Nat. Acad. Sci. USA, vol. 75, pp. 2602–2610, elementary transition in Shaker potassium channels,” Science, vol. 264, 1978. pp. 578–582, 1994. [3] M. Noda, S. Shimizu, T. Tanabe, T. Takai, T. Kayano, T. Ikeda, H. Taka- [30] G. Yellen, “The moving parts of voltage-gated ion channels,” Q. Rev. hashi, H. Nakayama, Y. Kanaoka, and N. Minamino, “Primary struc- Biophys., vol. 31, pp. 239–295, 1998. ture of electrophorus electricus sodium channel deduced from cDNA [31] F. Bezanilla, “The voltage sensor in voltage-dependent channels,” sequence,” Nature, vol. 312, pp. 121–127, 1984. Physiol. Rev., vol. 80, pp. 555–592, 2000. [4] T. M. Tempel, D. M. Papazian, T. L. Schwarz, Y. N. Jan, and L. Y. [32] Y. Jiang, A. Lee, J. Chen, V.Ruta, M. Cadene, B. T. Chait, and R. MacK- Jan, “Sequence of a probable potassium channel component encoded innon, “X-ray structure of a voltage-dependent K channel,” Nature, at Shaker locus in drosophila,” Science, vol. 237, pp. 770–775, 1987. vol. 423, pp. 33–41, 2003. [5] A. Parsegian, “Energy of an ion crossing a low dielectric membrane: [33] Y. Jiang, V. Ruta, J. Chen, A. Lee, and R. MacKinnon, “The principle of gating charge movement in a voltage-dependent K channel,” Nature, Solutions to four relevant electrostatic problems,” Nature, vol. 221, pp. vol. 423, pp. 42–48, 2003. 844–846, 1969. [34] B. E. Cohen, M. Grabe, and L. Y. Jan, “Answers and questions from [6] IEEE Trans. Nanobiosci. (Special Issue on Ion Channels), vol. 4, no. 1, KvAP structure,” Neuron, vol. 39, pp. 395–400, 2003. pp. 1–9, Mar. 2005. [35] S. R. Durell and H. R. Guy, “Atomic scale structure and functional [7] O. P. Hamill, A. Marty, E. Neher, B. Sackmann, and F. J. Sigworth, models of voltage-gated potassium channels,” Biophys. J., vol. 62, pp. “Improved patch-clamp techniques for high-resolution current recording 238–250, 1992. from cells and cell-free membrane patches,” Pflugers Arch., vol. 391, pp. [36] W. A. Caterall, “Molecular properties of voltage-sensitive sodium chan- 85–100, 1981. nels,” Annu. Rev. Biochem., vol. 55, pp. 953–985, 1986. [8] C. M. Armstrong and F. Bezanilla, “Currents related to movement of the [37] C. S. Gandhi and E. Y. Isacoff, “Molecular models of voltage sensing,” gating particles of the sodium channels,” Nature, vol. 242, pp. 459–461, J. Gen. Physiol., vol. 120, pp. 455–463, 2002. 1973. [38] C. A. Ahern and R. Horn, “Specificity of charge-carrying residues in the [9] R. D. Keynes and E. Rojas, “Kinetics and steady-state properties of the voltage sensor of potassium channels,” J. Gen. Physiol., vol. 123, pp. charged system controlling sodium conductance in the squid giant axon,” 205–216, 2004. J.Physiol. (London), vol. 239, pp. 393–434, 1974. [39] S. R. Durell, I. H. Shrivastava, and H. R. Guy, “Models of the structure [10] C. A. Vandenberg and F. Bezanilla, “A sodium channel model of gating and voltage-gating mechanism of the Shaker K channel,” Biophys. J., based on single channel, macroscopic ionic and gating currents in the vol. 87, pp. 2116–2130, 2004. squid giant axon,” Biophys. J., vol. 60, pp. 1511–1533, 1991. [40] F. Bezanilla, “Perspective: Voltage sensor movements,” J. Gen. Physiol., [11] E. Perozo, R. MacKinnon, F. Bezanilla, and E. Stefani, “Gating cur- vol. 120, pp. 465–473, 2002. rents from a nonconducting mutant reveal open-closed conformations [41] D. M. Starace and F. Bezanilla, “A proton pore in a potassium channel in Shaker K channels,” Neuron, vol. 11, pp. 353–358, 1993. voltage sensor reveals a focused electric field,” Nature, vol. 427, pp. [12] W. N. Zagotta, T. Hoshi, J. Dittman, and R. Aldrich, “Shaker potassium 548–552, 2004. channel gating III: Evaluation of kinetic models for activation,” J. Gen. [42] F. Bezanilla and B. Roux, “Mechanism of voltage sensor operation in Physiol., vol. 103, pp. 321–362, 1994. voltage-dependent channels,” Biophys. J., vol. 86, pp. 522a–522a, 2004. [13] F. Bezanilla, E. Perozo, and E. Stefani, “Gating of Shaker K channels. [43] S. A. Monks, D. J. Needleman, and C. Miller, “Helical structure and II. The components of gating currents and a model of channel activa- packing orientation of the S2 segment in the Shaker K channel,” J. tion,” Biophys. J., vol. 66, pp. 1011–1021, 1994. Gen. Physiol., vol. 113, pp. 415–423, 1999. [14] N. E. Schoppa and F. J. Sigworth, “Activation of Shaker potassium [44] Y. Li-Smerin, D. H. Hackos, and K. J. Swartz, “Alpha-helical structural channels. III. An activation gating model for wild-type and V2 mutant elements within the voltage-sensing region domains of a K channel,” channel,” J. Gen. Physiol., vol. 111, pp. 313–342, 1998. J. Gen. Physiol., vol. 115, pp. 33–50, 2000. [15] D. Sigg, H. Qian, and F. Bezanilla, “Kramers’ diffusion theory applied [45] L. G. Cuello, M. Cortes, and E. Perozo, “Molecular architecture of the to gating kinetics of voltage dependent channels,” Biophys. J., vol. 76, KvAP voltage dependent K channel in a lipid bilayer,” Science, vol. pp. 782–803, 1999. 306, pp. 491–495, 2004. 48 IEEE TRANSACTIONS ON NANOBIOSCIENCE, VOL. 4, NO. 1, MARCH 2005

[46] M. Laine, M. C. Lin, J. P. Bannister, W. R. Silverman, A. F. Mock, B. [72] C. M. Armstrong and F. Bezanilla, “Inactivation of the sodium channel. Roux, and D. M. Papazian, “Atomic proximity between S4 segment II. Gating current experiments,” J. Gen. Physiol., vol. 70, pp. 567–590, and pore domain in Shaker potassium channels,” Neuron, vol. 39, pp. 1977. 467–481, 2003. [73] T. Hoshi, W. N. Zagotta, and R. W. Aldrich, “Biophysical and molec- [47] A. Cha, G. Snyder, P. R. Selvin, and F. Bezanilla, “Atomic scale move- ular mechanisms of Shaker potassium channel inactivation,” Science, ment of the voltage-sensing region in a potassium channel measured via vol. 250, pp. 533–538, 1990. spectroscopy,” Nature, vol. 402, pp. 809–813, 1999. [74] A. Cha, P. C. Ruben, A. L. George, E. Fujimoto, and F. Bezanilla, [48] J. Richardson, D. M. Starace, F. Bezanilla, and A. M. Correa, “Scanning “Voltage sensors in domains III and IV, but not I and II, are immobilized NaChBac topology using LRET,” Biophys. J., 2005, to be published. by Na channel fast inactivation,” Neuron, vol. 22, pp. 73–87, 1999. [49] K. S. Glauner, L. M. Mannuzzu, C. S. Gandhi, and E. Y. Isacoff, “Spec- [75] B. Chanda and F. Bezanilla, “Tracking voltage-dependent conforma- troscopic mapping of voltage sensor movements in the Shaker potassium tional changes in skeletal muscle sodium channel during activation,” J. channel,” Nature, vol. 402, pp. 813–817, 1999. Gen. Physiol., vol. 120, pp. 629–645, 2002. [50] R. O. Blaustein, P. A. Cole, C. Williams, and C. Miller, “Tethered [76] C. R. Cantor and P. R. Schimmel, Biophysical Chemistry. Part II. Tech- blockers as molecular ’tape measures’ for a voltage-gated K channel,” niques for the Study of Biological Structure and Function. New York: Nat. Struct. Biol., vol. 7, pp. 309–311, 2000. Freeman, 1980. [77] P. R. Selvin, “Lanthanide-based resonance energy transfer,” IEEE J. Se- [51] Q.-X. Jiang, D.-N. Wang, and R. MacKinnon, “Electron microscopic lect. Topics Quantum Electron., vol. 2, no. 4, pp. 1077–1087, Dec. 1996. analysis of KvAP voltage dependent K channel in an open conforma- [78] C. Gonzalez, E. Rosenman, F. Bezanilla, O. Alvarez, and R. Latorre, tion,” Nature, vol. 430, pp. 806–810, 2004. “Periodic perturbations in Shaker K channel gating kinetics by dele- [52] J. M. Fernandez, F. Bezanilla, and R. E. Taylor, “Effect of chloroform tions in the S3–S4 linker,” Proc. Nat. Acad. Sci., vol. 98, pp. 9617–9623, on the movement of charges within the nerve membrane,” Nature, vol. 2001. 297, pp. 150–152, 1982. [79] D. M. Starace, P. R. Selvin, and F. Bezanilla, “Resonance energy transfer [53] Santacruz-Toloza, L. Y. Huang, S. A. John, and D. M. Papazian, “Gly- measurements on transmembrane motion of Shaker K channel voltage cosylation of Shaker potassium channel protein in insect cell culture and sensing region,” Biophys. J., vol. 82, p. 174a, 2002. in xenopus oocytes,” Biochemistry, vol. 33, pp. 5607–5613, 1994. [80] B. Chanda, O. K. Asamoah, B. Roux, and F. Bezanilla, “Measuring the [54] O. K. Asamoah, B. Chanda, and F. Bezanilla, “A spectroscopic survey transmembrane movement of the voltage sensor in the Shaker potassium of gating-induced environmental changes in the Shaker potassium channel,” Biophys. J. Abstracts, to be published. channel,” Biophys. J., vol. 86, p. 432a, 2004. [81] D. J. Posson, C. Miller, F. Bezanilla, and P. R. Selvin, “Luminescence [55] N. Yang and R. Horn, “Evidence for voltage-dependent S4 movement in resonance energy transfer studies of the Shaker K voltage-sensor using sodium channels,” Neuron, vol. 15, pp. 213–218, 1995. fluorescent scorpion-toxin acceptor,” Biophys. J. Abstracts, 2005, to be [56] N. Yang, A. L. George, and R. Horn, “Molecular basis of charge move- published. ment in voltage-gated sodium channels,” Neuron, vol. 16, pp. 113–122, [82] W. Treptow, B. Maigret, C. Chipot, and M. Tarek, “Coupled motions 1996. between the pore and voltage-sensor domains: A model for Shaker B, a [57] O. S. Baker, H. P. Larsson, L. M. Mannuzzu, and E. Y. Isacoff, “Three voltage-gated potassium channel,” Biophys. J., vol. 87, pp. 2365–2379, transmembrane conformation and sequence-dependent displacement 2004. of the S4 domain in Shaker K channel gating,” Neuron, vol. 20, pp. [83] C. A. Ahern and R. Horn, “Stirring up controversy with a voltage sensor 1283–1294, 1998. paddle,” Trends Neurosci., vol. 27, no. 6, pp. 303–307, 2004. [58] H. P. Larsson, O. S. Baker, D. S. Dhillon, and E. Y. Isacoff, “Transmem- [84] K. J. Swartz, “Towards a structural view of gating in potassium chan- brane movement of the Shaker K channel S4,” Neuron, vol. 16, pp. nels,” Nature Rev. Neurosci., vol. 5, pp. 905–916, 2004. 387–397, 1996. [85] E. Perozo, M. Cortes, and L. G. Cuello, “Structural rearrangements un- [59] S. P. Yusaf, D. Wray, and A. Sivaprasadarao, “Measurement of the derlying K -channel activation gating,” Science, vol. 285, pp. 73–78. movement of the S4 segment during activation of a voltage-gated [86] Y. Jiang, A. Lee, J. Chen, M. Cadene, B. T. Chait, and R. MacK- potassium channel,” Pflugers Arch. Eur. J. Physiol., vol. 433, pp. innon, “Crystal structure and mechanism of a calcium-gated potassium 91–97, 1996. channel,” Nature, vol. 417, pp. 515–522. [60] D. M. Starace, E. Stefani, and F. Bezanilla, “Voltage-dependent proton [87] , “The open pore conformation of potassium channels,” Nature, vol. transport by the voltage sensor of the Shaker K channel,” Neuron, vol. 417, pp. 523–526, 2002. [88] S. M. Webster, D. Del Camino, J. P. Dekker, and G. Yellen, “Intracel- 19, pp. 1319–1327, 1997. lular gate opening in Shaker K channels defined by high-affinity metal [61] D. M. Starace and F. Bezanilla, “Histidine scanning mutagenesis of basic bridges,” Nature, vol. 428, pp. 864–868, 2004. residues of the S4 segment of the Shaker K channel,” J. Gen. Physiol., [89] F. Bezanilla and E. Perozo, “The voltage sensor and the gate in ion chan- vol. 117, pp. 469–490, 2001. nels,” in Advances in Protein Chemistry, D. Rees, Ed. New York: El- [62] L. D. Islas and F. J. Sigworth, “Electrostatics and the gating pore of sevier, 2003, vol. 63. Shaker potassium channels,” J. Gen. Physiol., vol. 117, pp. 69–89, 2001. [90] A. Sonnleitner, L. M. Mannuzzu, S. Terakawa, and E. Y. Isacoff, “Struc- [63] O. K. Asamoah, J. P. Wuskell, L. M. Loew, and F. Bezanilla, “A fluoro- tural rearrangements in single ion channels detected optically in living metric approach to local electric field measurements in a voltage-gated cells,” Proc. Nat. Acad. Sci. USA, vol. 99, no. 20, pp. 12 759–12 764, ion channel,” Neuron., vol. 37, pp. 85–97, 2003. 2002. [64] D. M. Papazian, X. M. Shao, S.-A. Seoh, A. F. Mock, Y. Huang, and D. [91] R. Blunck and F. Bezanilla, “Fluorescence recordings of a low number H. Weinstock, “Electrostatic interactions of S4 voltage sensor in Shaker of voltage gated K channels,” Biophys. J., vol. 82, p. 267a, 2002. K channel,” Neuron, vol. 14, pp. 1293–1301, 1995. [92] V. Borisenko, T. Lougheed, J. Hesse, E. Fureder-Kitzmuller, N. Fertig, [65] S. K. Tiwari-Woodruff, C. T. Schulteis, A. F. Mock, and D. M. Pa- J. C. Behrends, G. A. Woolley, and G. J. Schutz, “Simultaneous optical pazian, “Electrostatic interactions between transmembrane segments and electrical recording of single gramicidin channels,” Biophys J., vol. mediate folding of Shaker K channel subunits,” Biophys. J., vol. 72, 84, no. 1, pp. 612–622, 2003. pp. 1489–1500, 199. [93] R. Blunck, J. L. Vazquez-Ibar, Y. S. Liu, E. Perozo, and F. Bezanilla, [66] L. M. Mannuzzu, M. M. Moronne, and E. Y. Isacoff, “Direct phys- “Fluorescence measurements of KcsA channels in artificial bilayers,” ical measure of conformational rearrangement underlying potassium Biophys. J., pt. 2, vol. 84, no. 2, pp. 124a–125a, 2003. channel gating,” Science, vol. 271, pp. 213–216, 1996. [67] A. Cha and F. Bezanilla, “Characterizing voltage-dependent conforma- tional changes in the Shaker K channel with fluorescence,” Neuron, vol. 19, pp. 1127–1140, 1997. Francisco Bezanilla received the Ph.D. degree from [68] , “Structural implications of fluorescence quenching in the Shaker the Catholic University, Santiago, Chile, in 1968. K channel,” J. Gen. Physiol., vol. 112, pp. 391–408, 1998. He held postdoctoral positions at the National [69] R. Blunck, D. M. Starace, A. M. Correa, and F. Bezanilla, “Detecting Institutes of Health, Bethesda, MD, and at the rearrangements of Shaker and NaChBac in real-time with fluorescence University of Rochester, Rochester, NY. He held spectroscopy in patch-clamped mammalian cells,” Biophys. J., vol. 86, a position as a professor in The University of pp. 3966–3980, 2004. Chile. He is currently the Hagiwara Professor of [70] A. Kuzmenkin, F. Bezanilla, and A. M. Correa, “Gating of the bacte- Neuroscience in the Departments of Physiology rial sodium channel NaChBac: Volatgedependent charge movement and and Anesthesiology at the D. Geffen School of gating currents,” J. Gen. Physiol., vol. 124, pp. 349–356, 2004. Medicine, University of California, Los Angeles. [71] E. Loots and E. Y. Isacoff, “Protein rearrangements underlying slow in- His primary research is aimed at the understanding activation of the Shaker K channel,” J. Gen. Physiol., vol. 112, pp. of the structure–function relation of ion channels and voltage-dependent 377–389, 1998. membrane processes.