Free polyunsaturated fatty acids (PUFAs) modulate the voltage dependence of voltage-gated ion channels. As an important consequence thereof, PUFAs can suppress epileptic seizures and cardiac arrhythmia. However, molecular details for the interaction between PUFA and ion channels are not well understood. In this study, we have localized the site of action for PUFAs on the voltage-gated Shaker K channel by introducing positive charges on the channel surface, which potentiated the PUFA effect. Furthermore, we found that PUFA mainly affects the final voltage sensor movement, which is closely linked to channel opening, and that specific charges at the extracellular end of the voltage sensor are critical for the PUFA effect. Because different voltage-gated K channels have different charge profiles, this implies channel-specific PUFA effects. The identified site and the pharmacological mechanism will potentially be very useful in future drug design of small-molecule compounds specifically targeting neuronal and cardiac excitability.
Polyunsaturated fatty acids (PUFAs) are components of phospholipids in the cell membrane, where they contribute to the fluidity and directly affect the activity of membrane proteins such as voltage-gated ion channels (Schmidt et al., 2006; Börjesson and Elinder, 2008; Y. Xu et al., 2008). In addition, free PUFAs play important physiological roles by affecting different membrane proteins, including ion channels (Boland and Drzewiecki, 2008; Sfondouris et al., 2008), and beneficial effects of PUFAs on heart arrhythmias and epilepsy have been reported (Lefevre and Aronson, 2000; Leaf et al., 2003). We suggested previously that PUFAs are important active substances in the fat-rich ketogenic diet used to treat severe epilepsy, by acting on voltage-gated K (Kv) channels (X.P. Xu et al., 2008). Specifically, PUFAs shift the voltage dependence of activation of the Shaker Kv channel via an electrostatic mechanism (Börjesson et al., 2008, 2010) (schematized in Fig. 1, A and B). Small shifts can have surprisingly large effects on excitability; a −5-mV shift is equivalent to increasing the number of K channels by a factor of 3 in the frog myelinated axon (Börjesson et al., 2010). The charge of the PUFA head group determines the direction of the effect, which has been referred to as the lipoelectric mechanism (Börjesson et al., 2008, 2010). However, because the site of action of PUFA on voltage-gated ion channels is unknown, the actual molecular mechanism of action for PUFA was hitherto unclear.
PUFAs partition into the cell membrane and likely interact with the channel at several positions, but the most likely target for the lipoelectric effect is the voltage sensor itself. In this study, we first set out to identify the site of action for PUFAs on Kv channels (referred to as the “PUFA action site” throughout). This is of crucial importance for understanding the mechanism by which PUFAs affect voltage-gated ion channels and for understanding why different channels are differently sensitive to PUFAs. A centrally located ion-conducting pore domain of the voltage-gated ion channel is surrounded by four voltage sensor domains (VSDs; Fig. 1 C) (Long et al., 2007; Börjesson and Elinder, 2008). Each VSD is composed of four transmembrane segments (S1–S4) (Long et al., 2007). S4 contains 4–10 positively charged residues responding to changes in the membrane voltage by sliding along negative charges in S1–S3, thereby turning the channel on or off (Tombola et al., 2006; Börjesson and Elinder, 2008). Several pharmacologically important sites have been identified in voltage-gated ion channels (Catterall et al., 2007; Börjesson and Elinder, 2008) (Fig. 1 C). These include: (a) Pore-blocking compounds (α-KTx for K channels), such as the scorpion charybdotoxin and agitoxins, which bind to the extracellular entrance of the ion-conducting pore; (b) quaternary ammonium compounds, such as local anaesthetics, antiarrhythmics, and antiepileptics, which bind to the cavity of the internal entrance to the ion-conducting pore; (c) lipophilic voltage sensor trapping toxins, such as spider hanatoxin, which bind in the lipid membrane close to the extracellular end of S3; and, finally, (d) several newly discovered small-molecule compounds, such as the antiepileptic drug retigabine, which bind to the pore domain, thereby keeping the intracellular gate open. A central question is if PUFAs act on any of these four sites. Because PUFAs affect many voltage-gated ion channels with different primary structures (Boland and Drzewiecki, 2008), the PUFA action site is likely a common favorable lipophilic environment rather than a high-affinity lock-and-key site. Sites 3 and 4 above are both accessible via the lipid bilayer and are thus likely candidates.
Second, to better understand the interaction between the PUFA and the Kv channel, we also aimed to identify which step in the channel activation chain is affected by PUFA. A Kv channel is activated in several steps; that is, several molecular rearrangements precede channel opening (Zagotta et al., 1994; Schoppa and Sigworth, 1998; Keynes and Elinder, 1999; Börjesson and Elinder, 2008). In response to a positive voltage inside the cell, the four S4s first move independently of each other, from the intracellular side of the channel to the extracellular side. Subsequently, a final conformational change in S4 occurs, involving cooperativity between the VSDs, followed by opening of the ion-conducting pore. In theory, PUFAs could induce channel opening by affecting one or several of those activation steps.
We found that PUFA molecules affecting Kv channel voltage dependence act close to the lipid-facing outer halves of S3 and S4, representing a novel pharmacological site of action distinct from those reported previously for toxins (Swartz, 2007; Kopljar et al., 2009) and potential antiepileptics (Xiong et al., 2007; Blom et al., 2009; Lange et al., 2009). We also report that PUFAs act on the final voltage sensor transition and that the effect critically depends on specific amino acid residues at the outer end of S4, suggesting a channel specificity in the PUFA effect on ion channels. This offers the prospect of designing new channel-modifying compounds potentially useful in the treatment of epilepsy and heart arrhythmia, and may help design novel dietary regiments for alleviating these diseases.
MATERIALS AND METHODS
Molecular biology and channel expression
For the electrophysiological experiments, we expressed the Shaker H4 channel in Xenopus laevis oocytes. Residues 6–46 were deleted to remove fast inactivation (Hoshi et al., 1990). This channel is referred to as Shaker WT-IR throughout. The conducting ILT mutation (V369I/I372L/S376T) (Smith-Maxwell et al., 1998) and the nonconducting ILT mutation (referred to as ILT/W434F; C245V/V369I/I372L/S376T/C426A/W434F) were provided by E. Isacoff (University of California, Berkeley, Berkeley, CA). Frog surgery, oocyte handling, and cRNA injection (0.05–0.5 ng/oocyte for Shaker point mutants, 12.5 ng/oocyte for ILT mutant, and 33 ng/oocyte for ILT/W434F mutant) followed the procedures described previously (Börjesson et al., 2010). Animal experiments were approved by the local Animal Care and Use Committee at Linköping University. Cysteine point mutations were introduced using QuikChange Site-Directed Mutagenesis kit (Agilent Technologies) and verified by sequencing. cRNA for point mutants and ILT mutants was prepared using the mMessage mMachine T7 kit (Invitrogen).
Ion and gating currents were measured 3–6 d after cRNA injection using the two-electrode voltage clamp technique (CA-1B amplifier; Dagan Corporation). Glass tips of borosilicate, filled with 3 M KCl, had a resistance of 0.5–2.0 MΩ. The holding voltage was −80 mV. Ion currents were measured at voltages between −80 and +60 mV (between +20 and +150 mV for the ILT mutant) in 5-mV increments. OFF gating currents of ILT/W434F were measured by first stepping to −120 mV (100 ms), then to prepulse voltages between −120 and +10 mV in 5-mV increments (100 ms), and finally to −100 mV (200 ms). All experiments were performed at room temperature (∼20°C). The amplifier’s leak and capacitive compensation were used, and the currents were low-passed filtered at 5 kHz. The control solution contained (in mM): 88 NaCl, 1 KCl, 15 HEPES, 0.4 CaCl2, and 0.8 MgCl2. pH was adjusted to 7.4, with NaOH yielding a final sodium concentration of ∼100 mM. 4,7,10,13,16,19-all-cis-docosahexaenoic acid (DHA) was prepared, stored, and applied as described previously (Börjesson et al., 2008). DHA concentrations mentioned in this work are the effective concentrations, i.e., 70% of the nominal concentration (quantified in Börjesson et al., 2008).
A PUFA bound to the channel has an apparent pKa value close to 7.4 (Börjesson et al., 2008). An alteration in the local pH close to the active PUFA alters the effect on the channel’s voltage sensitivity by altering the proportion of deprotonated charged PUFA (Fig. 2 C; see also triple mutant in Börjesson et al., 2008). The local pH in turn depends on the fixed charges close to the PUFA. To alter the charge of the substituted cysteines, we used the cysteine-specific thiol reagents 2-aminoethyl methanethiosulfonate hydrochloride (MTSEA+), [2-(trimethylammonium)ethyl] methanethiosulfonate bromide (MTSET+), and sodium [2-sulfonatoethyl] methanethiosulfonate (MTSES−; Toronto Research Chemicals). The MTS reagents were applied continuously to the bath solution using a gravity-driven perfusion system. The reagents were applied until full modification was achieved (typically 100 µM MTSEA+, 100 µM MTSET+, or 1 mM MTSES− for 200 s). The modification was assayed functionally in two-electrode voltage-clamped oocytes. The main substance used to introduce a positive charge at each mutated cysteine was MTSEA+. The reason for using the smaller MTSEA+ instead of the larger MTSET+ is that we wanted a more defined charge localization and less steric effects than what is obtained from the larger reagents. MTSEA+ can, in contrast to MTSET+ and MTSES−, pass the cell membrane (Holmgren et al., 1996) and was therefore first tested on Shaker WT-IR to identify possible interference from MTSEA+ in the cytoplasm. 100 µM MTSEA+ applied for >480 s had no effect on Shaker WT-IR (Table I). MTSEA+ experiments were done at pH 7.4 because a higher pH would already from the beginning push the PUFAs toward a deprotonated state, and no further effect will be seen upon modification.
In addition to possible altered PUFA sensitivity upon MTSEA+ modification, MTSEA+ modification per se of a residue that is close to the voltage sensor can also induce electric and/or steric effects on channel activation. The electric and steric components for a charged MTS reagent can be derived by using differently charged MTS reagents (Broomand and Elinder, 2008). By also testing the effect of the negatively charged MTSES−, the electric versus steric components can be quantified. The steric component is calculated as ΔVsteric = (ΔVMTSES + ΔVMTSEA)/2, and the electrostatic component is calculated as ΔVelectric = (ΔVMTSEA − ΔVMTSES)/2.
Analysis of data
The K conductance GK(V) was calculated as
where IK is the steady-state current at the end of a 100-ms pulse, V is the absolute membrane voltage, and Vrev is the reversal potential for the K channel, set to −80 mV. The voltage dependence analysis and quantification of G-V shifts were done by measuring the shift at the 10% level as described previously (Börjesson et al., 2008, 2010) to avoid interference from endogenous currents and block by intracellular ions, as sometimes seen at the most positive voltages. For illustrative reasons, the figures presented here were generated by fitting a Boltzmann curve raised to the fourth power to the conductance data
where A is the amplitude of the curve, and V1/2 and s are the midpoint and the slope, respectively. s was constrained to a common value in each panel in the figures. After this fit, the data were normalized to A = 1. However, this procedure only marginally affected the amplitudes (<10%).
Q-V was analyzed by integrating the OFF gating current before and after DHA treatment. The gating charge was normalized and plotted against the prepulse voltage, and the control curve slid along the voltage axis until it overlapped the DHA curve.
If a charge is located at the border between a low dielectric medium (the lipid membrane) and a high dielectric medium (water), the potential ψ(r) at the distance r from an elementary charge e0 (e.g., the PUFA charge) can be calculated with a modified Coulomb’s law (McLaughlin, 1989; Elinder and Århem, 2003),
where ε0 is the permittivity of free space (8.85 × 10−12 Fm−1), εa is the relative dielectric constant of the aqueous phase (80), and κ is the inverse of the Debye length in the aqueous phase (9.8 Å in the 1-K solution; see e.g., Elinder et al., 2001b). Eq. 3 was used to get a rough estimate of the distance between the PUFA carboxyl charge and the gating charges in S4 based on DHA-induced Q-V and G-V shifts for the ILT mutant. The calculation is based on the assumptions that the surface of the channel is smooth, and that the top S4 charges pop up on the channel’s surface one by one, leaving the interior of the channel unaffected with positive gating charges pairing with negative counter charges (see Fig. 6 A). This simple model has been evaluated previously (Elinder et al., 2001a) and proven surprisingly accurate (Elinder et al., 2001a; Broomand and Elinder, 2008).
Molecular K channel structure
The crystal structure of the Shaker channel is not determined. Therefore, we used the structure of the Kv1.2/2.1 chimera channel (Long et al., 2007) with Shaker side chains (provided by E. Lindahl, Royal Institute of Technology, Stockholm, Sweden) for the structural evaluations. The Kv1.2/2.1 chimera shares high sequence identity with the Shaker channel and has been shown previously to serve as an accurate Shaker model (Tao et al., 2010).
For the computer simulations of K channel gating, we used a six-state model (Fig. 5 A). C0 to C4 indicate states of the channel with 0–4 activated voltage sensors (S4). The movements of the voltage sensors are independent of each other. From state C4, the channel can open via a final cooperative step that involves all four subunits. The steady-state conditions are determined by α/β and by γ/δ, where
Vαβ is the voltage where α = β, Vγδ is the voltage where γ = δ, and zαβ and zγδ are the number of elementary charges moving through the membrane when the channel goes between two closed states or between C4 and O, respectively. Vαβ was set to −40 mV for WT-IR and −80 mV for ILT. Vγδ was set to −40 mV for WT-IR and +100 mV for ILT. zαβ was set to 2 for both WT-IR and for ILT. zγδ was set to 1 for both WT-IR and for ILT. G-V and the gating charge versus voltage curve, Q-V, can easily be computed (see Fig. 5, F and G):
Average values are expressed as mean ± SEM. When comparing DHA-induced shifts of mutants with WT-IR, one-way ANOVA together with Dunnett’s multiple comparison test was used. One-way ANOVA followed by Bonferroni’s multiple comparison test was used to compare DHA sensitivity for each mutant before and after MTS modification. Mean values for MTS-induced shifts were analyzed using a two-tailed one sample t test, where mean values were compared with a hypothetical value of 0. P < 0.05 is considered as significant.
Cysteine mutations in the VSD but not in the pore domain affect the PUFA sensitivity
During previous investigations on the voltage-gated Shaker K channel, we found that some cysteine point mutations reduced the potency of the ω-3 PUFA DHA to affect the channel’s voltage dependence compared with the Shaker WT-IR. Here, we have used DHA as the test substance but will refer to the effects as “PUFA effects” because we have found almost identical Shaker channel effects for six different investigated PUFAs, in sharp contrast to the lack of effects for monounsaturated, saturated, and methyl esters of fatty acids (Börjesson et al., 2008). In a previous investigation, we found that low concentrations such as 2 µM DHA had clear effects on the voltage dependence of the Shaker K channel (Börjesson et al., 2008), and that this small effect is expected to have substantial effects on excitability (X.P. Xu et al., 2008; Börjesson et al., 2010). In the present work, we aimed at detailing the interaction between DHA and the channel; therefore, much higher DHA concentrations (7–70 µM) were used.
We hypothesized that mutagenesis of residues close to the PUFA action site affects PUFA sensitivity more than distant residues do. Therefore, to localize this site, we performed a systematic cysteine scan of a large surface of the channel based on the localization of the initial cysteine mutants. The point-mutated Shaker channels were then expressed in Xenopus oocytes and studied with two-electrode voltage clamp. 17 residues were selected for mutagenesis. The investigated residues are distributed over different segments of the channel (Fig. 2 A) and cover the lipophilic surfaces of the extracellular halves of S3, S4, S5, and S6 (Fig. 2 B, red residues). The reason for this selection is that PUFAs are lipophilic substances and thus prone to interact with residues facing a lipid environment, and they should be relatively close to the gating charges in the voltage sensor (Börjesson et al., 2008) (Fig. 2 B, blue sticks).
The ability of 70 µM DHA at pH 7.4 to shift the conductance versus voltage, G-V, curve along the voltage axis (as shown in Fig. 1 A) was tested for each cysteine mutant. Out of the 17 residues investigated, all but four mutants showed no alteration in DHA sensitivity (<1.6-mV deviation from the sensitivity of WT-IR; Table I). The four mutants with larger alterations in DHA sensitivity (>±2.8 mV) compared with WT-IR were I325C, T329C, I360C, and L366C. Residues I325C and T329C, located in the outer half of helix S3, had lost their sensitivity and were not significantly affected by DHA. I360C (located in S4) also showed reduced DHA sensitivity, but it was not statistically different from WT-IR in the one-way ANOVA analysis. L366C (in S4) displayed increased DHA sensitivity compared with WT-IR. The alterations in DHA sensitivity were not correlated with the minor alterations in the channels’ voltage sensitivity caused by the mutation per se, denoted in Table I as the voltage for the 10% level of the G-V curve. These initial experiments suggested that residues in the VSD but not in the pore domain are important for PUFA acting on the channel’s voltage dependence.
Positive charges close to the PUFA action site increase the PUFA potency
If the cysteine substitution of a residue induces only marginal changes in the side chain properties, the mutation will probably not alter the PUFA effect. Such cysteine mutants would thus incorrectly be interpreted as distant to the PUFA action site. Therefore, to further explore the critical area for PUFA action, we introduced a charge to each of the investigated residues. This charge can have three different effects on PUFA efficiency. (1) A positive charge close to the PUFA action site will increase the local pH and consequently deprotonate the PUFA molecule (Börjesson et al., 2008). This in turn will increase the PUFA effect on the channel’s voltage dependence (Fig. 2 C). (2) A positive charge close to the PUFA action site will increase PUFA affinity and thereby increase PUFA effects at concentrations below saturating effects. (3) A positive charge located in the voltage sensor S4 moving toward the PUFA molecule promotes larger effects of the PUFA molecule on the voltage sensor movement. These three components can also work together, and we do not aim at quantifying the impact of each component. What is important is that in all three cases, high-impact residues should be close to the PUFA action site. A change in PUFA potency, independent of the underlying mechanism, is therefore our readout to indentify high-impact residues. To experimentally change the charge of each mutated cysteine, we modified the cysteine with the positively charged cysteine-specific reagent MTSEA+. This modification of a residue that is close to the voltage sensor will electrostatically shift the G-V curve in a positive direction along the voltage axis, in addition to any steric effect on channel activation. The closer to the voltage sensor, the larger the expected electrostatic effect is (Elinder et al., 2001a; Broomand and Elinder, 2008). For these experiments, we used pH 7.4 because the charge-induced effects are expected to be larger close to the apparent pKa value. Control experiments showed that WT-IR was not affected by MTSEA+ (Fig. 3 A).
Residues in the pore domain are close to the voltage sensor but distant from the PUFA action site
Residue 416 is located at the extracellular end of S5 at atomic distance to the top charge of S4 (residue R362) in the open state (Broomand et al., 2003; Lainé et al., 2003). 70 µM DHA at pH 7.4 shifts the G-V curve for F416C with −3.5 mV (Fig. 3 B, left, and Table I), which is close to what has been found for WT-IR (Fig. 3 A, left, and Table I). When modifying F416C with MTSEA+, the G-V curve was shifted +27 mV (Fig. 3 B, middle, and Table I), suggesting that F416 is very close to the voltage sensor if the induced shift is of electrostatic origin. To estimate the electrostatic component of this shift, we also modified F416C with the negatively charged MTSES− (Elinder et al., 2001a). The electrostatic component was 11.6 mV (see Materials and methods for calculations), which means that a charge at position 416 electrostatically affects the voltage sensitivity with 11.6 mV, suggesting that residue 416 is located a short distance (∼10 Å; Eq. 3) from the voltage sensor, as shown previously (Broomand et al., 2003; Lainé et al., 2003). To find out if 416 is also close to the PUFA action site, we tested if the PUFA-induced shift was enhanced by the MTSEA+ modification. The PUFA-induced shift was not significantly different from the unmodified channel or the WT-IR channel (Fig. 3 B, right, and Table I). Three similar examples are S421C, G452C, and V453C, all located in the pore domain at the top of S5 and S6, respectively (Table I). This is consistent with the top of S5 and S6 being close to the voltage sensor S4, but it also suggests that the pore domain is distant from the PUFA action site. One of the residues we investigated in the pore domain, G420C did not show any alterations (MTSEA+, MTSES−, or changes in DHA response). Either it is buried from the extracellular solution and thus not modified, or it is pointing away from both the voltage sensor and the PUFA action site.
Residues in S3 and S4 are both close to the voltage sensor and to the PUFA action site
Of 11 investigated residues in S3–S4, 10 were modified by either MTSEA+ or MTSES− as judged either directly by a change in the channel’s voltage sensitivity or indirectly by a change in the channel’s DHA sensitivity (Table I). Only residue I364C was not affected by the MTS reagents, suggesting a buried position not exposed to the extracellular solution. 4 of the 10 modified residues (I325C, T329C, A359C, and I360C) increased their sensitivity to DHA when the residues were made positively charged by MTSEA+.
For I360C, located two residues outside the first gating charge in S4, and for I325C, in S3, 70 µM DHA at pH 7.4 shifts the G-V with −1.5 and −1.1 mV, respectively (Fig. 3, C and D, left, and Table I). For both mutations, modification with MTSEA+ clearly shifts the G-V with +7 and +19 mV, respectively (Fig. 3, C and D, middle, and Table I). The MTSEA+ modification significantly increased the DHA-induced shift to −8.5 and −5.9 mV, respectively (Fig. 3, C and D, right, and Table I), clearly more than for WT-IR. Therefore, both residues 325 and 360 are suggested to be located in or close to the voltage sensor (as also found in the structure) and to the PUFA action site. MTSEA+ modification of T329C in S3 and A359C in S4 also significantly affected the voltage sensitivity of the channel and the PUFA-induced shift (Table I). This is consistent with the cysteine mutagenesis data (see above); the three cysteine mutations that significantly altered the DHA sensitivity (325C, 329C, and 366C) are geographically close to or overlap with the residues that increase their DHA sensitivity when charge modified (325C, 329C, 359C, and 360C). Examination of the molecular structure showed that high-impact residues are clustered in a small region of the lipid-facing S3–S4 corner of the VSD (Fig. 4 A). An estimation of the PUFA action site locates the PUFA carboxyl charge in the lipid bilayer adjacent to S3 and S4 (Fig. 4, A and B). Intriguingly, this is the same motif on the channel as membrane phospholipids are suggested to interact with to alter channel gating (Milescu et al., 2009).
Strategy to analyze which step in the activation chain is affected by PUFA
So far, we have localized the PUFA action site to the periphery of the Shaker K channel close to S3 and S4, distant from the pore region. This also suggests that PUFA affects the movement of S4, rather than movements in the pore domain associated with the channel opening. However, it is not clear which steps in the activation chain are affected. A simple five-step model can be used in the analysis of the channel opening (Fig. 5 A). It is generally agreed that the activation can be divided into two main components. (1) The first component is independent outward movements of the four S4 helices (corresponding to the four steps from C0 to C4). In reality, each S4 moves in several steps (Zagotta et al., 1994; Keynes and Elinder, 1998b; Schoppa and Sigworth, 1998; DeCaen et al., 2009), but for simplicity, we fuse these early transitions to a single step in each subunit. (2) When all S4s are in an activated position (Fig. 5 A, C4), the channel can open via a final voltage-dependent step in which all S4s move together (Sigg et al., 1994; Zagotta et al., 1994; Keynes and Elinder, 1998a; Schoppa and Sigworth, 1998; Ledwell and Aldrich, 1999; Pathak et al., 2005). The final S4 movement affects the pore domain to open the channel (Pathak et al., 2005). All steps in the model are voltage dependent, but most of the voltage dependence comes from the early steps (representing 80–85% of the gating charges compared with only 15–20% in the final step; Schoppa et al., 1992). The PUFAs could affect the voltage dependence of the opening either by affecting only the early steps, only the final step, or both types of steps.
In a previous investigation (Börjesson et al., 2008), we found larger PUFA effects on the ion currents than on the gating currents (mainly generated by the C0–C4 steps), suggesting that the opening step is more affected than the early transitions. However, the exact relation was difficult to resolve because it depends on the type of model used for Kv channel gating. To better quantify the PUFA effect on the different activation steps, we used the ILT mutant Shaker channel where three hydrophobic residues in S4 are substituted for three other hydrophobic residues (V369I, I372L, and S376T). The ILT mutation separates the early steps from the final step, which are then possible to study in isolation (Smith-Maxwell et al., 1998). The early transitions occur at voltages around −80 mV, whereas the opening occurs around +50 mV (see Fig. 5 B). Thus, the G-V in the ILT mutant only reports on the opening step, with no involvement of the early transitions. If the W434F pore mutation is introduced, occluding the pore with no effects on the voltage sensor movement (Perozo et al., 1993), we instead get information of the early steps in isolation. The ILT experiments were performed at pH 9.0 to make sure that the fatty acid is fully charged and to induce as large PUFA effects as possible for robust quantification; the pKa value for a fatty acid incorporated in the membrane is ∼7.4 (Rooney et al., 1983; Börjesson et al., 2008).
PUFA mainly affects the final opening step, with minor effects on the early transitions
First, we investigated the ion-conducting ILT mutant, which directly reports on the opening of the channel. DHA had a large effect on this K current (Fig. 5 C). 70 µM DHA at pH 9.0 increased the current at +60 mV by a factor of ∼10. DHA clearly shifted the G-V curve in the negative direction along the voltage axis (Fig. 5 D), qualitatively similar as the DHA-induced G-V shift for WT-IR. However, the shift was larger for ILT than for WT-IR. At pH 9.0, 70 µM DHA shifted G-V for ILT with −30 ± 2.5 mV (n = 5), whereas G-V for WT-IR was shifted −18 ± 1.4 mV (wild-type data from Börjesson et al., 2008). The larger effect on the G-V for the ILT mutant than for the WT-IR channel suggests that the opening step is more affected than the early steps. In the WT-IR channel, a large effect on the opening step is probably masked by much smaller effects on the early steps, which kinetically interfere with the opening step.
Second, to quantify the effects on the early steps, we also studied gating currents in the ILT/W434F mutant. DHA had a much smaller effect on the gating currents, Q-V, than on the conductance (Fig. 5, E and H). At pH 9.0, 70 µM DHA shifted Q-V for ILT/W434F with −5.0 ± 0.8 mV (n = 3; Fig. 5 E). Even though this effect on the gating currents is much smaller than that for the opening step, it suggests that the early transitions are also affected in the Shaker K channel. Fig. 5 H summarizes shift data for G-V and Q-V for both WT-IR and ILT (open bars). The smaller effect on G-V and the larger effect on Q-V for the WT-IR channel compared with ILT suggest that, under the assumption that the ILT mutation does not alter the relation of the DHA effects on the early and the late steps, the opening step and the major gating charge steps kinetically affect each other in WT-IR.
A simple gating model is quantitatively consistent with the experimental data
To quantitatively evaluate the experimental DHA data for the two activation steps, we used a simple gating model (Fig. 5 A and Eqs. 4–7). To mimic the effects of 70 µM DHA at pH 9.0, we shifted the voltage dependence of the first transitions (Vαβ for C0–C4) with −5 mV, as experimentally found for Q-V for ILT/W434F, and the voltage dependence for the opening step (Vγδ) with −30 mV, as experimentally found for the G-V shift for the ILT mutant (Fig. 5, F and G, for computations). The effect on the voltage dependence of the computed Q-V and G-V curves for wild-type and the ILT mutant induced by these shifts was subsequently measured. A summary of all experimental and computational shifts is shown in Fig. 5 H. The concordance between experimental and computational data is striking, suggesting that DHA affects the first activation steps with −5 mV and the final opening step with −30 mV.
Horizontal charge movement of R1 explains the differential effects on the early and the final transitions
A helical-screw motion of S4 (Catterall, 1986; Guy and Seetharamulu, 1986; Broomand and Elinder, 2008) suggests that the positive gating charges emerge one by one on the channel’s surface during activation (Keynes and Elinder, 1999; Elinder et al., 2001a). In the closed state C0, none of the gating charges in S4 are exposed on the surface, whereas in the open state O, the three outermost arginines, R1–R3, are exposed to the extracellular solution (Fig. 6 A) (Baker et al., 1998; Keynes and Elinder, 1999; Long et al., 2007; Börjesson and Elinder, 2008). In the last closed state C4, with all S4s in an activated but nonconducting position, we suggest that R1 and R2 are exposed to the extracellular solution with its charges at a distance of 16 Å away from the approximate PUFA position in the lipid bilayer adjacent to S3 and S4 (Fig. 6 B, green and orange residues; note that the color coding in Fig. 6 A corresponds to the color coding for the suggested position of R1 in Fig. 6 B). It has been suggested that S4’s final transition opens the channel (Keynes and Elinder, 1998a; Pathak et al., 2005; Phillips and Swartz, 2010). Assuming that the final position of S4 is according to the Kv1.2/2.1 chimera crystal structure, the charge of R1 moves to a position only 6 Å away from the PUFA when the channel is open (Fig. 6 B, red residue).
Can this structural hypothesis quantitatively explain our data? In previous work, we have successfully and with high accuracy used electrostatic equations (Eq. 3 in the present work) to determine distances at the surface of the Shaker K channel (Elinder et al., 2001a; Broomand and Elinder, 2008). Eq. 3 gives the change in surface potential as a function of distance from a charge, and it was used here to estimate the distance between the PUFA carboxyl charge and S4. Of course, the PUFA carboxyl charge will electrostatically interact with all charges in the channel. However, in a helical screw–like model, each step in the activation pathway is equivalent to removing the bottom charge of S4 and adding an S4 charge at the extracellular surface (Fig. 6 A, dashed arrows). Assuming a helical-screw motion of S4, the only electrostatic effect of PUFA on the channel, for each step in the activation pathway, is the interaction of PUFA with the added S4 charge on the extracellular surface (under the assumption that the interaction with the removed bottom S4 charge is negligible). All other PUFA–channel electrostatic interactions cancel each other because each S4 charge is replaced by the next S4 charge in a helical-screw model, and conserved negative counter ions always pair up with the gating charges within the electric field. Based on these assumptions, the 5-mV effect on the first S4 steps suggests that the PUFA charge is located 15.2 Å away from the position where the positive charges emerge on the channel protein’s surface. This is very close to the structural prediction of 16 Å. The 30-mV effect on the opening step suggests that the PUFA charge is located 6.3 Å away from R1 in the open state, which again is almost identical to our prediction from the structural model. Collectively, our experimental data combined with the presented calculations suggest that R1 and R2 are exposed to the extracellular solution at a distance of ∼16 Å from the PUFA action site in the final closed state, C4. In the opening step (from C4 to O), R1 translates along the longitudinal axis of S4 and swings around the S4 helix toward the head groups of the lipid bilayer, close to the location of the charge of the bound PUFA (∼6 Å).
Charge mutations in S4 suggest channel-specific effects of PUFA on channel gating
The above mentioned structural model, where R1 moves from a position ∼16 Å away from the PUFA charge to a position only ∼6 Å away during the opening step, implies that the positive charge at R1 is critical for the effect of PUFAs. If the PUFAs interact electrostatically with R1 to promote opening, a negative charge at R1 would instead be expected to reduce or even reverse the PUFA effect. To alter the charge of R1 without altering the size of the side chain, we changed the arginine to a cysteine and modified with differently charged MTS reagents (Fig. 6 C): negatively charged MTSES− (mutant called R362−; Fig. 7 B) and positively charged MTSET+ (mutant called R362+; Fig. 7 C). These experiments were performed at pH 9.0 to promote a negative charge of the carboxyl group of DHA, avoiding changes in the charge of the DHA (pKa = 7.4; Börjesson et al., 2008) as a result of possible changes in local pH caused by the differently charged MTS reagents. In the same investigation, we showed that introducing three positive charges close to the voltage sensor only affected the local pH with 0.3 pH units.
7 µM DHA at pH 9.0 reduced, rather than increased, the current at −35 mV for R362− and induced a small but significant G-V shift in depolarizing direction (+2.3 ± 0.6 mV; n = 5; Fig. 7 B). This experiment suggests that DHA, despite the electrostatic repulsion from the negatively charged 362 residue, finds it attractive to bind to a site close to S3/S4. In contrast, R362+ restored the PUFA sensitivity (Fig. 7 C), giving similar G-V shifts as for WT-IR (Fig. 7 A). The structural model also suggests that an additional gating charge at position 359 (one turn above R1, also called R0) could further potentiate the PUFA effect by providing an extra charge for the PUFA to interact with. Indeed, when attaching MTSET+ to A359C (A359+) at pH 9.0, the DHA-induced G-V shift was increased from −11 to −18 mV (Fig. 7 D). For this channel, 2.1 µM DHA in its deprotonated form shifted the G-V with as much as −6.4 ± 0.4 mV (n = 3), reinforcing the idea that a channel with a positive charge at R0 can be very much affected by physiological concentrations of DHA. Fig. 7 E summarizes the charge-modifying data for R1 and R0, suggesting that PUFAs influence the voltage dependence of the opening step mainly by interacting electrostatically with the outer end of S4. The experiments changing the charge of R0 and R1 (a) support the proposed localization of the PUFA action site, (b) suggest that it is a final S4 transition moving the outermost gating charge toward the lipid bilayer that underlies the actual opening step, and (c) suggest that PUFAs should have very different effects on different ion channels depending on the presence of a charge at positions R0 and/or R1. This last insight highlights the selective effects on different ion channels and opens up the development of future medical drugs selectively affecting neuronal and cardiac excitability.
In this study, we have localized the site of action for PUFAs affecting the voltage dependence of the Shaker K channel to a position in the lipid bilayer close to the extracellular halves of S3 and S4 in the VSD. We have also found that DHA mainly affects the final S4 transition, which is closely linked to the channel opening.
Several models for S4 motion are consistent with the PUFA data
A tilted helical-screw motion in three steps (Keynes and Elinder, 1999; Tiwari-Woodruff et al., 2000; Lecar et al., 2003) is consistent with all of our experimental data (Fig. 8 A, top), but variations on the same theme could also explain the data: a helical-screw motion followed by a tilt of S4 in the final step (Pathak et al., 2005) (Fig. 8 A, middle), and a translational motion of a 310 helix followed by a 310-to-α-helix conversion in the final step (Shafrir et al., 2008; Bjelkmar et al., 2009; Schow et al., 2010) (Fig. 8 A, bottom). A small electrostatic PUFA effect on the early S4 movements (C0 to C4) can easily be understood from a simple surface charge theory, in which the PUFA carboxyl group provides additional fixed negative charges on the extracellular surface. If the charge is located close to the VSD, the modified transmembrane field will be sensed by the voltage sensor, and a smaller depolarizing step is needed to open the channel (Elinder and Århem, 2003). The mechanism is the same for all models. The larger PUFA effect on the opening step can be explained in a similar way if voltage sensor charges are moving closer to the PUFA action site during the opening step (C4→O). This occurs in all described models (Fig. 8 A). A gating charge movement in the plane of the membrane is also supported by the finding that a charge at R1 and R0 has profound effects on the PUFA potency. Such a model also implies the possibility that the PUFA affinity might be state dependent: a higher affinity in the open state than in a closed state. However, this has not been tested in the present investigation. These data are in concordance with fluorescence measurements suggesting that S4 moves laterally, with only a smaller transmembrane component, during the opening step (Pathak et al., 2005). The exact VSD rearrangement underlying the opening step is not known; therefore, it is difficult to differentiate between the models in Fig. 8 A. However, a recent investigation shows that the top charge of S4, R1, moves in an outward direction from the membrane during the final transition (Phillips and Swartz, 2010), thus supporting the helical-screw model in the top panel of Fig. 8 A.
A new site and a new mechanism
Does the PUFA action site reported in this investigation overlap with the site for other compounds (Fig. 1 C), and is the mechanism similar? We conclude that the PUFA action site (Fig. 8 B, orange) does not overlap with binding sites for other compounds. We also conclude that the mechanism of action for the PUFAs is unique. Kv channel openers are rare, but shifts in the voltage dependence similar to those induced by PUFAs are seen for a few synthetic compounds (for example, retigabine, ZnPy, and acrylamide on the Kv7.2/3 channel). However, these substances stabilize the open-channel conformation through interactions with distinct sites in the pore domain (Fig. 8 B, yellow residues) (Xiong et al., 2007; Blom et al., 2009; Lange et al., 2009) and not through interactions with the VSD. Also, the ciguatera toxin gambierol interferes with Kv channel gating (Kopljar et al., 2009) by binding to the lipid-exposed surface of the pore domain and stabilizing the closed channel.
Other compounds bind to the VSD to trap the channel in a certain conformation. Voltage sensor–trapping toxins bind to the outer part of S3 and the S3–S4 linker (Fig. 8 B, red residues) and trap the channel in an open or closed conformation depending on which conformation provides the highest toxin affinity (Swartz, 2007; Börjesson and Elinder, 2008). Lipophilic toxins and PUFAs demonstrate the diversity of molecules that can access the channel via the surrounding membrane. NH29 binds in a water-exposed crevice in the VSD and stabilizes the open channel through interactions with S2 and S4 (Peretz et al., 2010). The voltage sensor–trapping toxin site is close to the PUFA action site (the NH29 site is slightly more distant), but the mechanism for trapping is suggested to be of steric origin and thus different from the electrostatic PUFA mechanism. To our knowledge, this is the first report of a gating-modulating drug that electrostatically affects the voltage sensor to promote the final opening step. The beauty of the mechanism is that both reduced and increased excitability can be reached depending on the charge of the substance (Börjesson et al., 2008, 2010).
We would like to stress that the PUFA action site reported in this work was identified by studying the ability of PUFAs to change Shaker channel voltage dependence and is therefore the site of action for PUFAs acting by the lipoelectric mechanism. This does not exclude the presence of other PUFA–channel interfaces affecting Kv channels in other ways. Interestingly, a second PUFA site of action on Kv channels was recently identified in the ion-conducting pore (Decher et al., 2010). PUFA binding to this site induces fast channel inactivation and thereby K current inhibition. In earlier work, we have noted PUFA-induced Shaker channel inactivation induced by the application of arachidonyl amine (Börjesson et al., 2010) and high (≥210 µM) concentrations of PUFA (Börjesson et al., 2008). The Shaker channel possesses the critical isoleucine for pore block (Decher et al., 2010), suggesting that the inactivation previously reported by us may be explained by compound binding to the second site where the effect is not dependent on the charge of the compound.
The critical role of the charge of residues close to PUFAs suggests that different channels may be differently sensitive to PUFAs depending on their charge profile. Channels with several positive charges at the outer half of S3 and S4 may potentially be more PUFA sensitive. For instance, Kv2.1 lacks R1 but instead possesses two arginines further extracellular at positions corresponding to 358 and 359 in Shaker, having the potential to enhance the PUFA effect. Channels like Kv4.1, Kv4.2, and Kv11.1 lacking R1 could, on the contrary, be less PUFA sensitive. In support of this prediction, the PUFA arachidonic acid does not shift the voltage dependence of neither the activation nor the steady-state inactivation of Kv4.2 (Villarroel and Schwarz, 1996). Both the location of the PUFA site of action and the electrostatic mechanism on the opening step provide an interesting mechanistic basis for future development of small-molecule compounds to modulate electrical excitability to target epilepsy and cardiac arrhythmia.
We thank Lucas Cleve for some of the experiments; Björn Wallner and Erik Lindahl for the Kv1.2/2.1-based Shaker channel structure; Ehud Isacoff for the ILT and ILT/W434F clones; and Peter Larsson, Anders Blomqvist, and Stefan Thor for comments on the manuscript.
This work was supported by the Swedish Research Council, the Swedish Heart-Lung Foundation, the Swedish Brain Foundation, the County Council of Östergötland, Queen Silvia’s Anniversary Foundation, King Gustaf V and Queen Victoria’s Freemasons Foundation, Stina and Birger Johansson’s Foundation, and the Swedish Society for Medical Research.
Kenton J. Swartz served as editor.
2-aminoethyl methanethiosulfonate hydrochloride
sodium [2-sulfonatoethyl] methanethiosulfonate
[2-(trimethylammonium)ethyl] methanethiosulfonate bromide
polyunsaturated fatty acid
voltage sensor domain