Considerable controversy surrounds the location of the closed channel gate in members of the Cys-loop receptor family of neurotransmitter-gated ion channels that includes the GABAA, glycine, acetylcholine, and 5-HT3 receptors. Cysteine-accessibility studies concluded that the gate is near the cytoplasmic end of the channel in acetylcholine and GABAA receptors but in the middle of the 5-HT3A receptor channel. Zn2+ accessibility studies in a chimeric 5-HT3-ACh receptor suggested the gate is near the channel's cytoplasmic end. In the 4-Å resolution structure of the acetylcholine receptor closed state determined by cryoelectron microscopy, the narrowest region, inferred to be the gate, is in the channel's midsection from 9' to 14' but the M1–M2 loop residues at the channel's cytoplasmic end were not resolved in that structure. We used blocker trapping experiments with picrotoxin, a GABAA receptor open channel blocker, to determine whether a gate exists at a position more extracellular than the picrotoxin binding site, which is in the vicinity of α1Val257 (2') near the channel's cytoplasmic end. We show that picrotoxin can be trapped in the channel after removal of GABA. By using the state-dependent accessibility of engineered cysteines as reporters for the channel's structural state we infer that after GABA washout, with picrotoxin trapped in the channel, the channel appears to be in the closed state. We infer that a gate exists between the picrotoxin binding site and the channel's extracellular end, consistent with a closed channel gate in the middle of the channel. Given the homology with acetylcholine and 5-HT3 receptors there is probably a similar gate in those channels as well. This does not preclude the existence of an additional gate at a more cytoplasmic location.
In the closed, resting state, ion conduction through the Cys-loop receptor neurotransmitter-gated ion channels is blocked by a gate (or gates) in the transmembrane, channel-forming domain. Agonist binding to the extracellular domain induces a propagated conformational change that opens the closed state gate, allowing ion conduction through the channel (Karlin, 2002; Auerbach, 2003; Absalom et al., 2004; Akabas, 2004; Lester et al., 2004). In the continued presence of agonist, the channels enter nonconducting desensitized state(s). The conformations of the channels are different in the closed and desensitized states, and the gate(s) that block ion conduction in these kinetically distinct states may be different (Auerbach and Akk, 1998; Williams and Akabas, 1999; Wilson and Karlin, 2001; Purohit and Grosman, 2006). To understand the conformational changes that occur as the Cys-loop receptor channels open it is important to know the location of the closed channel gate(s). Evidence from a variety of experiments has led to differing conclusions regarding the location of the closed channel gate.
The ion channel in the Cys-loop receptors is largely lined by residues from the five M2 membrane-spanning segments that form an inner ring of α helices (Imoto et al., 1988; Leonard et al., 1988; Akabas et al., 1994; Xu and Akabas, 1996; Reeves et al., 2001; Miyazawa et al., 2003; Goren et al., 2004). The other three transmembrane segments, M1, M3, and M4, form an outer ring of helices that separate the M2 helices from the lipid bilayer (Blanton and Cohen, 1992; Miyazawa et al., 2003; Unwin, 2005). To facilitate comparisons between superfamily members, residues in the channel-lining M2 segment are referred to by an index numbering system (Miller, 1989); 0' is the absolutely conserved positively charged residue at the cytoplasmic end of M2, aligned with GABAA receptor α1Arg255, and 20' is the extracellular ring of charge, aligned with α1Asn275 (Imoto et al., 1988). The M2 segment appears to be α helical from −1' to 27' (Bera et al., 2002; Miyazawa et al., 2003).
Determining the location of the closed channel gate has been a subject of considerable controversy. Using the substituted cysteine accessibility method (SCAM), we and others showed that when applied extracellularly in the absence of agonist, charged sulfhydryl-reactive reagents reacted with cysteines substituted at the 2' level in the acetylcholine (ACh) and GABAA receptors. The gate was inferred to be below the 2' level, at the channel's cytoplasmic end (Akabas et al., 1994; Xu and Akabas, 1996). The conclusions of these studies regarding the location of the channel gate were limited by the fact that the sulfhydryl reagent reaction rates were not measured. To address this shortcoming in the ACh receptor, the reaction rates of methanethiosulfonate (MTS) reagents with channel-lining, M2 segment–substituted cysteines were measured in the absence and presence of agonist (Pascual and Karlin, 1998; Wilson and Karlin, 1998). At some positions and for some reagents, rates differed by 1,000-fold, whereas at certain adjacent positions they only differed by a few fold. Based on these results and on differences in the reaction rates of sulfhydryl reagents applied from the extracellular and cytoplasmic sides in the absence of ACh, it was concluded that the gate was located near the cytoplasmic end of M2 (Wilson and Karlin, 1998). Two potential caveats to the interpretation of these results are the possibility that the cysteine mutations may increase the spontaneous open probability and/or that some of the sulfhydryl reagents might act as weak, partial agonists, thereby allowing reaction to occur in the open rather than the closed state (Akabas, 2004). A subsequent SCAM study in the 5HT3A receptor attempted to address this issue by applying the sulfhydryl reagents in the presence of a high affinity competitive antagonist. They concluded that the gate was in the middle of the channel, below the 14' level (Panicker et al., 2002). In contrast, in a chimeric ACh-5HT3 receptor using a related approach, substituting histidines for M2 residues and measuring Zn2+ block, it was concluded that the gate was near the cytoplasmic end of the channel (Paas et al., 2005). The problem of the mutations altering the spontaneous open probability was not addressed in this work and some of the records show evidence of significant increases in spontaneous open probability with inhibition of baseline current by Zn2+ before the addition of agonist. Biochemical studies in the ACh receptor showed photoincorporation of a noncompetitive antagonist, 3-(trifluoromethyl)-3-(m-[(125)I] iodophenyl)diazirine, in the 9' and 13' region in the absence of agonist but not in the presence of agonist. This was inferred to be a hydrophobic environment in the closed state, perhaps consistent with a channel gate (White and Cohen, 1992). Finally, in the 4-Å resolution structure of the ACh receptor closed state that was solved using cyroelectron microscopic images of two- dimensional helical crystals, the narrowest region in the channel was a constriction, ∼6 Å in diameter, in the midsection from 9' to 14' that was inferred to be the channel gate, “a hydrophobic girdle” (Miyazawa et al., 2003). It should be noted that the 4-Å resolution structure failed to resolve the residues in the M1–M2 loop, the gate location inferred by other studies. Molecular dynamics simulations based on the ACh receptor structure have suggested that the 9' to 14' region may form a hydrophobic gate (Amiri et al., 2005; Corry, 2006). Thus, disparate methods have led to divergent conclusions regarding the position of the closed channel gate in Cys-loop receptors.
In the present work we sought to use an alternative approach to define the position of the channel gate in GABAA receptors. We have used open channel blocker trapping experiments to define the gate location relative to a known channel blocker binding site. Blocker trapping was used previously to determine the location of the channel gate in potassium channels (Armstrong, 1971; Holmgren et al., 1997). We used picrotoxin, a GABAA receptor open channel blocker that binds near the 2' position near the cytoplasmic end of the channel. We hypothesized that if picrotoxin was trapped in the channel after the removal of GABA by closing of a more extracellular channel gate, then the currents elicited by a subsequent GABA application should appear to activate more slowly than before picrotoxin application because picrotoxin's off-rate is slower than the GABA-induced channel opening rate (Newland and Cull-Candy, 1992; Maconochie et al., 1994). It would then be necessary to demonstrate that the picrotoxin-trapped state of the channel in the absence of GABA is the closed state and not an open-blocked or desensitized state. We used the state-dependent reactivity of various engineered cysteine residues as reporters for the channel's conformational state to show that the picrotoxin-trapped state is likely to be a closed state.
Picrotoxin inhibition of GABAA receptors is complex but most consistent with a noncompetitive, use-dependent, open channel block mechanism (Newland and Cull-Candy, 1992; Yoon et al., 1993; Chang and Weiss, 2002; Muroi et al., 2006; Olsen, 2006). Picrotoxin is an uncharged, rigid, roughly spherical molecule that is ∼9 Å in diameter (Xu et al., 1995). Picrotoxin binds at a site that is only accessible from the extracellular side in the open state (Newland and Cull-Candy, 1992). Several lines of experimental evidence support the location of a picrotoxin binding site in the GABAA receptor channel at the 2' level. First, 2' mutations alter picrotoxin block (Ffrench-Constant et al., 1993; Zhang et al., 1994). Second, picrotoxin protects a cysteine substituted for α1Val257 (2') from modification by charged sulfhydryl reagents applied extracellularly without protecting cysteines substituted at more extracellular channel lining positions (Xu et al., 1995). Third, the β3 2' mutation A252V increases GABA affinity but decreases picrotoxin affinity (Buhr et al., 2001). Finally, phenylalanine substitutions at the 6' level in any subunit eliminate picrotoxin block, presumably by sterically blocking access to the more cytoplasmic 2' binding site (Gurley et al., 1995). Thus, picrotoxin appears to be a GABAA receptor open channel blocker that binds in the channel lumen at the 2' level. We present additional experimental support for picrotoxin binding in the channel at the 2' level. It may also bind at a second allosteric site outside the channel (Yoon et al., 1993) and we will present evidence consistent with this and demonstrate that this interaction only decreases GABA-induced currents by 20–30%, discriminating it from the complete block caused by binding within the channel.
Materials And Methods
In Vitro Transcription and Oocyte Expression
The rat GABAA receptor subunit cDNAs in the pGEMHE vector were linearized with NheI for in vitro transcription using T7 RNA polymerase (Promega) (Goren et al., 2004). mRNA was stored at −80°C in diethylpyrocarbamate-treated water at a concentration of 200 μg/ml. Stage V–VI Xenopus laevis oocytes were defolliculated by treatment with 2 mg/ml collagenase Type 1A (Sigma-Aldrich) for 65 min. They were washed in OR2 (82.5 mM NaCl, 2 mM KCl, 1 mM MgCl2, 5 mM HEPES, pH 7.5 with NaOH) and kept in OR3 (70% Leibovitz L-15 medium [Invitrogen] supplemented with 10 mM HEPES, 50 μg/ml tetracycline, and 50 μg/ml gentamicin). Oocytes were injected 24 h after isolation with 50 nl of a 1:1:1 mixture of α1:β2:γ2S subunit mRNA and were kept in OR3 medium for 2–4 d at 17°C. Mutant subunit mRNA was substituted for wild-type subunits where necessary.
The electrophysiological recordings were conducted at room temperature in an ∼250-μl chamber continuously perfused at a rate of 5–6 ml/min with nominally calcium-free frog ringer (CFFR) composed of 115 mM NaCl, 2.5 mM KCl, 1.8 mM MgCl2, 10 mM HEPES, pH 7.5 with NaOH. Currents were recorded from Xenopus oocytes using two electrode voltage clamp recording at a holding potential of −60 mV. The ground electrode was connected to the bath via a 3 M KCl/agar bridge. Glass microelectrodes filled with 3 M KCl had a resistance of <2 MΩ. Data were acquired and analyzed using a TEV-200 amplifier (Dagan Instruments), a Digidata 1322A data interface (Axon Instruments), and pClamp 8 software (Axon Instruments). Currents were elicited by 10–20-s applications of GABA separated by at least 4 min of CFFR wash to allow complete recovery from desensitization. Currents were judged to be stable if the variation between consecutive GABA pulses was <10%. GABA, picrotoxin, and penicillin were obtained from Sigma-Aldrich.
Curve Fitting and Data Analysis
The time course of currents elicited by GABA was fitted using a mono- or biexponential function using the Chebyshev algorithm implemented in Clampfit (Axon Instruments). Data are presented as mean ± SEM.
Picrotoxin Application Slows the Subsequent Channel Opening Rate
We expressed wild-type rat α1β2γ2S GABAA receptors in Xenopus oocytes. Prior to picrotoxin application, the time course of the current increase due to application of 10 μM GABA (EC50) could be fit with a single exponential function. The average rise time was 340 ± 75 ms (n = 6) (Fig. 1, A and B, control). Coapplication of 100 μM picrotoxin and 10 μM GABA almost completely inhibited the GABA-induced current (Fig. 1 A, trace 2). After inhibition by picrotoxin, GABA was washed out, presumably allowing the channels to close, and then picrotoxin was washed out for 5 min. Upon reapplication of GABA, a double exponential function was required to fit the current increase with rise times of 400 ± 85 ms and 4200 ± 700 ms (n = 6). The amplitude of slow component was 76 ± 4% and of the fast component 24 ± 4% (n = 12). The longer rise time was significantly slower than the initial rise time (two tailed Student's t test, P < 0.05) (Fig. 1 B, compare control and picrotoxin traces). We infer that the slower opening component represented the opening of channels in which picrotoxin had remained trapped during the washout period. The faster opening component activates with a time course similar to the activation before picrotoxin application. We assume that picrotoxin exited from some of the channels either as GABA was washed out or during the 5-min period after the picrotoxin washout. The presence of this component does not affect our results. The trapped picrotoxin could be washed out of the channels either by repeated 10–20-s EC50 GABA applications (Fig. 1 A, traces 3–5) or by a single prolonged application of saturating GABA (Fig. 1 E). After these GABA applications, the current rise could once again be fit with a single exponential function with a rise time of 370 ± 70 ms (n = 6) (Fig. 1 B, compare control and recovery traces). This demonstrated the expected reversibility of the picrotoxin block. Picrotoxin trapping can be repeated several times on the same oocyte (unpublished data).
Similar results were obtained using 200 μM GABA, a saturating concentration, for the test responses before and after picrotoxin trapping. The average rise time of the GABA-induced current before picrotoxin was faster, 150 ± 50 ms (n = 6). The fast rise time was limited by the solution exchange rate at the oocyte surface, a physical constraint imposed by the size of the cell. This does not affect our results as the rise time after picrotoxin application is significantly slower than the solution exchange time. After picrotoxin washout, the current increase required a bi-exponential fit with rise times of 150 ± 30 and 1200 ± 200 ms (n = 6) (Fig. 1, C and D).
Application of 100 μM picrotoxin in GABA's absence had no effect on the current rise time of subsequent GABA applications, consistent with picrotoxin being an open channel blocker, that is, the binding site was not accessible in the closed state (unpublished data). This is consistent with the results of Muroi et al. (2006) who showed that picrotoxin application in the absence of GABA did not alter the fluorescence intensity of site-specific fluorescent-labeled GABA receptors, whereas GABA activation did.
Penicillin, a Voltage-dependent, Open Channel Blocker, Prevents Picrotoxin Block
To use picrotoxin to determine the position of the gate we must establish that picrotoxin binds in the channel. While most evidence supports the hypothesis that picrotoxin binds in the channel, some experiments have suggested the existence of a second site. Therefore, we sought to provide additional evidence that the major component of picrotoxin block is mediated via binding in the channel.
Because picrotoxin is uncharged, picrotoxin block is not voltage dependent. In contrast, penicillin is an anion at physiological pH 7.5 and blocks GABA-induced currents in a noncompetitive, voltage-dependent manner (Chow and Mathers, 1986; Twyman et al., 1992; Fujimoto et al., 1995; Sugimoto et al., 2002). The amount of block by 1 mM penicillin was 67 ± 3% when measured at −60 mV and 79 ± 3% at 0 mV (n = 4) (unpublished data). The voltage dependence is consistent with the penicillin binding/blocking site being in the ion channel where it can be influenced by the transmembrane voltage.
Penicillin (10 mM) blocked ∼95% of the GABA-induced current but did not alter the subsequent rate of channel reopening (Fig. 2, A and B). We could not detect penicillin trapping in the closed channel. This may be due to the low affinity of penicillin for the channel and its high on and off rates (Chow and Mathers, 1986; Twyman et al., 1992). Alternatively, penicillin may bind at a position that prevents channel closing while penicillin is in the channel, as was seen with QX-222 in the ACh receptor channel (Neher and Steinbach, 1978).
When coapplied, 10 mM penicillin and 10 μM picrotoxin blocked the GABA-induced current (Fig. 2 A, trace 4) but after a 5-min wash, the subsequent channel opening rate was not slowed (Fig. 2, A and B, right trace). We infer that penicillin, with a higher on-rate, blocked the majority of the channels, preventing picrotoxin from binding. Thus, after washout of picrotoxin and penicillin from the bath, the subsequent channel opening rate was not affected by the prior presence of picrotoxin because in the presence of both penicillin and picrotoxin most of the channels had been blocked by penicillin and not by picrotoxin. This is consistent with a competitive interaction between penicillin and picrotoxin for inhibition of GABAA receptors. This interaction could arise because penicillin and picrotoxin bind at the same or overlapping sites. Alternatively, penicillin may bind at a more extracellular position in the channel and prevent picrotoxin from gaining access to a more cytoplasmically located binding site. We cannot distinguish between these possibilities but it does not alter the interpretation that this result provides support for the presence of a picrotoxin binding site in the channel lumen that mediates the picrotoxin trapping shown in Fig. 1.
Additional evidence to support a competitive interaction between penicillin and picrotoxin comes from sequential application and washout experiments. In the presence of GABA, 3 mM penicillin blocks ∼80% of the GABA current (Fig. 2 D). In the presence of GABA and penicillin, application of 10 μM picrotoxin, a concentration that by itself blocks ∼80% of the GABA-induced current (Fig. 2 C, right trace), blocked most of the remaining current (Fig. 2 D). Washout of penicillin, in the continuing presence of GABA and picrotoxin, resulted in a transient current increase that was blocked, presumably by picrotoxin still present in the bath (Fig. 2 D). The current appearing upon penicillin washout arose because the off-rate of penicillin is much faster than the on-rate for picrotoxin. This implies that picrotoxin could not bind to penicillin-blocked channels because if both picrotoxin and penicillin were simultaneously bound and inhibiting the GABAA receptors, then penicillin washout would not have resulted in a transient rise in the current. In contrast, once penicillin is removed, the channels that were blocked by penicillin are opened and can be blocked by picrotoxin, which is once again consistent with a competitive interaction most likely in the channel lumen.
Finally, we first applied a submaximal concentration of penicillin (300 μM) to block about one third of the GABA current and then applied 10 μM picrotoxin to block the remainder of the GABA current (Fig. 2 E, trace 2). 5 min after simultaneous washout of GABA, penicillin, and picrotoxin, the best fit to the rise of the GABA current was a biexponential function where the fast component accounted for about one third of the total current (Fig. 2 E, trace 3). This result is also consistent with a mutually exclusive interaction of penicillin and picrotoxin with GABAA receptors. Thus, in light of the voltage dependence of penicillin block, which implies that penicillin binds in the channel, and the competition between penicillin and picrotoxin that we have demonstrated, we infer that picrotoxin also binds in the channel lumen.
Picrotoxin Binds in the Channel Near the 2' Position
We previously showed that anionic sulfhydryl-reactive reagents reacted with an engineered cysteine at the 2' position in α1V257Cβ1γ2 receptors, inhibiting subsequent GABA-induced currents, whereas cationic reagents did not react (Xu and Akabas, 1996). The inhibition after modification of α1V257C by anionic sulfhydryl reagents precluded the possibility of determining whether the modification prevented picrotoxin binding. To circumvent this problem we used a double mutant, α1A254E/V257C with a cysteine at the 2' position and a glutamate at the adjacent −1' position, the amino acid found at the aligned position in the ACh receptor intermediate ring of charge (Imoto et al., 1988). In oocytes expressing α1A254E/V257Cβ2γ2, a 150-s coapplication of GABA and 2 mM MTSET+ caused a significant increase in the GABA-induced current during the application and resulted in a significant increase in the holding current after GABA washout (Fig. 3 A). We infer that in the presence of the engineered −1' glutamate, the cationic MTSET+ was able to penetrate sufficiently deep into the channel to allow reaction with the engineered 2' cysteine. The increased holding current after MTSET+ modification implies that the spontaneous open probability was increased in the MTSET+-modified channels. As can be seen in the figure (Fig. 3 A), the reaction went to completion during the 150-s application.
Prior to reaction with MTSET+, the α1A254E/V257Cβ2γ2 receptors were almost completely blocked by picrotoxin (Fig. 3 B) in a concentration-dependent manner (Fig. 3 E, filled circles). After reaction with MTSET+, the extent of inhibition by picrotoxin was significantly reduced but not eliminated (Fig. 3 D). The increased holding current was not inhibited but the remaining GABA-induced current was inhibited by ∼20% in a concentration-dependent manner (Fig. 3 E, open triangles). We infer that covalently modifying the 2' cysteine in one or both α1 subunits with the positively charged thioethyltrimethylammonium portion of MTSET+ significantly increases the Cl− conductance but prevents picrotoxin from binding in the channel at its channel-blocking, binding site. These data provide evidence, in addition to previously published results (Ffrench-Constant et al., 1993; Zhang et al., 1994; Gurley et al., 1995; Wang et al., 1995; Xu et al., 1995), to support the contention that picrotoxin binds in the GABAA receptor channel near the 2' level.
The 20% residual inhibition by picrotoxin after MTSET+ reaction could be due to incomplete reaction of MTSET+ with the channels but as we noted above, the MTSET+ reaction appeared to go to completion (Fig. 3 A). Some of the Cys could be modified by endogenous sulfhydryls or oxidized to higher order oxidation states that preclude reaction with MTSET+. Alternatively, the residual inhibition may be due to binding at another site on the GABAA receptor, outside of the channel, where inhibition is mediated by allosteric effects rather than by direct channel block. The existence of a second allosteric site outside the channel has been suggested by other investigators (Yoon et al., 1993).
In the Absence of GABA, the Picrotoxin-blocked Channel Appears To Be in the Close State and Not in the Desensitized State
Thus far we have provided evidence to support the hypothesis that picrotoxin binds within the GABAA receptor channel. Furthermore, we have shown the picrotoxin can be trapped in channels after the removal of GABA. To use picrotoxin trapping to locate the position of a closed channel gate we need to determine the conformational state of the picrotoxin-trapped (presumably agonist-free) channels. To accomplish this we used the state-dependent reactivity of two engineered cysteines, α1E303C and α1V280C, as reporters for the channel's conformational state. Previously, we showed that reaction rates at these positions were state dependent (Williams and Akabas, 1999, 2000, 2001; Bera et al., 2002). We chose these positions because in the structure of the homologous ACh receptor the aligned residues are separated by 20–30 Å and both are a similar distance from the picrotoxin binding site (Miyazawa et al., 2003). This separation is greater than the 9-Å diameter of picrotoxin so that picrotoxin could not simultaneously protect both positions.
The mutant α1E303C (in M3) fulfills two conditions that are important for this particular experiment. (1) pCMBS− does not react with α1E303C in the absence of GABA where the channels are presumably in the closed state (Williams and Akabas, 1999). (2) As we showed previously (Williams and Akabas, 1999), reaction with pCMBS− occurs in the presence of GABA because the subsequent GABA test currents are reduced (Fig. 4 A, compare the magnitude of trace 1 with traces 3 and 4). However, the reaction in the presence of GABA does not appear to occur in the open state because the current during the application of GABA + pCMBS is not altered compared with the current induced by GABA alone (Fig. 4 B). Note that the initial GABA-induced current and the current during the GABA + pCMBS− application can be superimposed (Fig. 4 B). This leads us to conclude that reaction of pCMBS− with α1E303C in the presence of GABA most likely occurs in a desensitized state but it cannot be the fast desensitized state from which channels frequently return to the open state. We infer that reaction occurs in a deeper desensitized state from which channels rarely return to the open state. Therefore, if pCMBS− reacts with α1E303C in the picrotoxin-trapped state it would suggest that this is a desensitized state.
Picrotoxin applied in the presence of GABA can be trapped in the α1E303Cβ1γ2S channels just as in wild- type channels. This trapping resulted in slowed channel opening after picrotoxin trapping and the need for a biexponential fit to the opening rate after picrotoxin was trapped (Fig. 4, C and D). We then examined the reactivity of pCMBS− with α1E303C in the picrotoxin-trapped state (Fig. 4 E). After two GABA test pulses (Fig. 4 E, traces 1 and 2), we coapplied picrotoxin and GABA and washed them out to create the picrotoxin-trapped state (Fig. 4 E, trace 3). Next we applied 500 μM pCMBS− for 1 min in the absence of GABA (Fig. 4 E, trace 4). After pCMBS− washout, we applied 200 μM GABA twice to allow complete escape of the trapped picrotoxin (Fig. 4 E, traces 5 and 6). After allowing time for recovery from desensitization, we applied a GABA test pulse (Fig. 4 E, trace 7), which had the same magnitude and time course as the initial GABA currents (trace 1). This implies that the cysteine at α1E303C was not modified during the pCMBS− application to the picrotoxin-trapped channels. To demonstrate that the cysteine was still available to react with pCMBS−, we next applied pCMBS− in the presence of GABA (Fig. 4 E, trace 8). The subsequent GABA test currents (Fig. 4 E, traces 9 and 10) were markedly reduced compared with the previous test current (Fig. 4 E, trace 7). This demonstrates that the cysteine did not react during the first pCMBS− application in the picrotoxin-trapped channels because it was available to react during the second pCMBS− application in the presence of GABA. Based on the fact that α1E303C does not react in the closed state but does react in an activated/desensitized state (Williams and Akabas, 1999), we infer that the picrotoxin-trapped state is not a desensitized state.
In the Absence of GABA, the Picrotoxin-blocked Channel Is in the Closed State
To provide further support for the hypothesis that the picrotoxin-trapped channels are in the closed state we used another cysteine mutant, α1V280C (M2 25'), that is located near the extracellular end of the channel-lining M2 segment. Reaction with MTSET+ significantly potentiated the subsequent GABA-induced currents (Fig. 5 B). The reaction rate of MTSET+ with this mutant was approximately seven times faster in the presence of GABA than in its absence (closed state rate 185 M−1s−1; open/activated state rate 1353 M−1s−1) (Fig. 5, A and B) (Bera et al., 2002). We could therefore use the MTSET+ reaction rate with α1V280C as a marker for the closed state. As with wild-type receptors, picrotoxin was trapped inside α1V280Cβ1γ2S receptors, leading to a slower opening rate after picrotoxin trapping that was best fit by a double exponential function (Fig. 5, C and D).
To determine whether the channels with picrotoxin trapped in them were in the open/activated or closed states, we examined the reactivity of α1V280C in the picrotoxin-trapped channels. We chose an MTSET+ concentration (50 μM) and a duration of application (30 s) that should react with >80% of open/activated-state channels but with <5% of closed-state channels (0.0015 M-s). After picrotoxin trapping (Fig. 5 E, trace 3), 50 μM MTSET+ was applied for 30 s (Fig. 5 E, trace 4). After application of 200 μM GABA to facilitate washout of the trapped picrotoxin (Fig. 5 E, trace 5), the subsequent GABA test currents were similar to the initial GABA test currents (Fig. 5 E, compare traces 1, 2 and 6, 7). To prove that the engineered cysteine was reactive with a higher concentration of MTSET+ we repeated the experiment in the same oocyte, but after picrotoxin trapping, we tested the effect of a 30-s application of 10 times higher concentration of MTSET+ (500 μM, 0.015 M-s) (Fig. 5 E, trace 9). After a 200 μM GABA application to wash out the trapped picrotoxin (trace not shown in figure between the two slashes), the subsequent GABA currents were significantly potentiated (Fig. 5 E, final two traces compared with initial two). These results imply that in the picrotoxin-trapped state the reactivity of α1V280C was consistent with the closed channel reaction rate. Lending further support to this hypothesis, a 30-s coapplication of 50 μM MTSET+ during washout of the trapped picrotoxin with a prolonged GABA application allowed reaction to proceed, consistent with the open/activated channel rate (Fig. 5 F). This resulted in the potentiation of the subsequent GABA test currents (Fig. 5 F, final two traces compared with initial two traces).
The goal of the present study was to use picrotoxin blocker trapping to determine whether a closed channel gate exists in the GABAA receptor between the picrotoxin binding site and the extracellular end of the channel. We have provided evidence that picrotoxin can be trapped in the GABAA receptor after removal of agonist. The existence of the picrotoxin-trapped state was demonstrated by the slowed rate of opening on the next application of GABA (Fig. 1). The slowing of GABAA receptor channel opening during the acquisition of a current–voltage relationship in the continuous presence of picrotoxin was previously noted (Newland and Cull-Candy, 1992). Considerable evidence in the literature supports the contention that picrotoxin binds in the GABAA receptor channel at the 2' level near α1V257 (Ffrench-Constant et al., 1993; Zhang et al., 1994; Gurley et al., 1995; Xu et al., 1995). We provided additional evidence to support this hypothesis through the penicillin competition experiments (Fig. 2) and by the inhibition of picrotoxin binding by MTSET+ modification of α1A254E/V257C channels (Fig. 3). Thus, in order to keep picrotoxin trapped in its binding site after the removal of agonist we infer that a gate must exist between the picrotoxin binding site at the 2' level and the extracellular solution because picrotoxin can only access its binding site from the extracellular side (Newland and Cull-Candy, 1992). We hypothesize that in the absence of agonist the picrotoxin-trapped state is likely to be either a closed state or a desensitized state, although we cannot exclude the possibility that it is a unique conformation of the receptor. We used the state-dependent reactivity of two cysteine substitution mutants, one in the M2 segment (α1V280C) and one in the M3 segment (α1E303C), in an attempt to distinguish whether the picrotoxin-trapped state resembles a closed or desensitized conformation. The α1V280C mutant reacted with MTSET+ at a rate consistent with reaction in the closed state and not in a GABA-activated state (Fig. 5). The α1E303C mutant did not react with pCMBS−, consistent with the picrotoxin-trapped channels being in the closed and not a desensitized state (Fig. 4). Thus, we infer that in the absence of agonist the picrotoxin-trapped state resembles the closed state. Furthermore, we infer that the gate that lies at a position extracellular to the picrotoxin binding site and that traps picrotoxin in its binding site is likely to be the closed state gate. This would be consistent with the narrow region between 9' and 14' observed in the 4-Å resolution ACh receptor structure forming a closed channel gate (Miyazawa et al., 2003; Unwin, 2005) and with the inference from SCAM experiments on the 5-HT3 receptor that the extracellular side of a closed channel gate was at the 14' level (Panicker et al., 2002) and with the inference from photolabeling experiments that the ACh receptor might have a gate at the 13' level (White and Cohen, 1992).
Further support for the existence of a closed channel gate between the extracellular end of the channel and the 2' level comes from the fact that picrotoxin is an open channel blocker. Picrotoxin blocks GABA currents only when it is applied from the extracellular side in the presence of GABA (Newland and Cull-Candy, 1992). It is unable to access its binding site in the closed state (Newland and Cull-Candy, 1992). Application of picrotoxin in the absence of GABA to GABAA receptors labeled with site-specific fluorophores caused no change in fluorescence, whereas activation by GABA in the absence of picrotoxin produced complex changes in fluorescence intensity (Muroi et al., 2006). Similar results were observed with GABAC receptors (Chang and Weiss, 2002). The inability of picrotoxin to access its binding site in the closed channel state is consistent with the presence of a closed channel gate between the extracellular end of the channel and the 2' level. An alternative explanation that we cannot exclude at present for picrotoxin not acting on the closed channel is that the binding site is only formed by a conformational change in the 2' region that occurs during channel activation. Our current results demonstrating picrotoxin trapping would argue against this alternative explanation.
Our current results do not preclude the possibility that there are additional gates at positions more cytoplasmic than the picrotoxin binding site. Picrotoxin is a roughly spherical molecule ∼9 Å in diameter. A constriction that would limit picrotoxin movement through the channel in the closed state might not be sufficient to limit the flux of small ions through the channel. In fact, the channel lumen presumably narrows below the 2' level to prevent picrotoxin from moving into the cytoplasmic channel vestibule. We think that picrotoxin does not move to a position below 2' because it would not have been able to protect α1V257C (2') from modification by extracellularly applied pCMBS− (Xu et al., 1995). Furthermore, we think that picrotoxin binding in the presence of GABA does not cause closure of a channel gate more extracellular than 6' because a cysteine substituted at 6', α1T261C, could react with extracellularly applied MTS reagents when the GABA-activated channels were blocked by picrotoxin (Xu et al., 1995).
Establishing the location of the picrotoxin binding site in the channel is critical for the interpretation of these experiments. As reviewed in the Introduction, considerable evidence indicates that a picrotoxin binding site is located near the cytoplasmic end of the GABAA receptor channel at the 2' level (Ffrench-Constant et al., 1993; Zhang et al., 1994; Gurley et al., 1995; Xu et al., 1995). We have provided additional evidence that picrotoxin binds in the channel lumen by showing the ability of penicillin to prevent picrotoxin binding (Fig. 2). Picrotoxin is uncharged and its block of GABA-induced currents is not voltage dependent (Newland and Cull-Candy, 1992). In contrast, penicillin is an anion at neutral pH and blocks GABAA receptors in a voltage- dependent manner consistent with it blocking by binding in the ion channel where it can experience the transmembrane voltage (Chow and Mathers, 1986; Twyman et al., 1992; Fujimoto et al., 1995; Sugimoto et al., 2002). Preapplication of penicillin to GABA-activated channels prevented subsequent inhibition by picrotoxin (Fig. 2). This implies that penicillin binds in the channel at and/or above the picrotoxin binding site and prevents picrotoxin from reaching its binding site from the extracellular end of the channel. This experiment strongly supports the hypothesis that there is a picrotoxin binding site in the channel. However, because the location of the penicillin binding site within the channel is not known it does not place additional constraints on the location of a picrotoxin binding site other than that it is in the channel.
To further support the location of a picrotoxin binding site near the M2 segment 2' level we demonstrated that a mutant channel that binds picrotoxin no longer binds picrotoxin after modification of a 2' cysteine. We previously showed that picrotoxin protected a Cys substituted for α1Val257 (2') from modification by extracellularly applied pCMBS− (Xu et al., 1995). Reaction of negatively charged sulfhydryl reagents with α1V257C caused significant inhibition of subsequent GABA-induced currents (Xu and Akabas, 1996) and this precluded investigating whether picrotoxin could still block the residual currents after modification. Furthermore, positively charged reagents did not react at this position (Xu and Akabas, 1996). In the current work, we have taken advantage of an observation that the cationic reagent MTSET+ can react with α1V257C (2') in the presence of a mutation at the adjacent −1' position, α1A254E, that inserts the amino acid present at the aligned position in the cation-selective Cys-loop family members. Picrotoxin blocked the unmodified double mutant α1A254E/V257C channels with similar affinity to wild-type channels (unpublished data). Thus, the mutations alone did not significantly alter the structure of the picrotoxin binding site. Reaction of MTSET+ with α1A254E/V257C in the presence of GABA had two effects: it caused a significant increase in the holding current and it increased the subsequent GABA-induced current. The increased holding current was most likely due to a large increase in the spontaneous open probability after MTSET+ modification. The increased GABA-induced currents were most likely due to a shift in the GABA EC50 of the modified channels. We cannot rule out the possibility that the increased currents also resulted from an increase in the single channel conductance but this would not affect the interpretation of the subsequent experiments with picrotoxin. In the MTSET+-modified channels, there is a major decrease in the extent of picrotoxin inhibition. We infer that modification of the 2' Cys by MTSET+ prevented picrotoxin from binding to its binding site within the channel lumen. This is unlikely to be an electrostatic interaction with picrotoxin because picrotoxin is uncharged. The inhibition of picrotoxin block could be due to direct steric interactions between the modified Cys and picrotoxin or it could be an indirect allosteric effect or electrostatic effect on the protein structure whereby modification of the 2' Cys alters a site at which picrotoxin normally binds. We believe that in light of the previously published result from our lab that picrotoxin protects the 2' Cys from modification by sulfhydryl reagents (Xu et al., 1995) and the mutagenesis results of others that show that 2' mutations prevent picrotoxin block (Ffrench-Constant et al., 1993; Zhang et al., 1995), the most likely interpretation of our result is that the MTSET+-modified 2' Cys prevents picrotoxin binding by a direct steric effect. Thus, we infer that the majority of the inhibition due to picrotoxin is due to binding at a site within the channel lumen.
There may be a second picrotoxin binding site on the GABAA receptor that accounts for the residual inhibition by picrotoxin after MTSET+ modification of the α1A254E/V257C channels. We cannot exclude the possibility that the residual inhibition by picrotoxin occurs in channels that did not react with MTSET+ although we think this is unlikely because reaction seems to have gone to completion. The residual inhibition is rapidly reversible (unpublished data). Work by others has suggested the existence of a second picrotoxin binding site besides the site within the channel (Yoon et al., 1993). Thus, based on the previously published results from our lab and from several others and on our current results, we conclude that picrotoxin binds at the 2' level in the GABAA receptor channel.
We have presented evidence that picrotoxin can be trapped for minutes in the GABAA receptor channel after washout of GABA and picrotoxin (Fig. 1). This implies that a gate or constriction exists in the channel between the picrotoxin binding site at the 2' level and the extracellular end of the channel. To infer that this is a closed state gate we need to establish the state of the GABAA receptor in the absence of GABA with picrotoxin trapped in the channel. We do not think that the channel can be in an open state because in the open state picrotoxin can leave its binding site. Thus, the channels are either in a desensitized state or a closed state, both of which are nonconducting states in which a gate or gates block ion conduction. There is extensive evidence using a variety of techniques including photoaffinity labeling, spectroscopy, cysteine accessibility, and kinetic analysis to indicate that in Cys-loop receptors the closed and desensitized states are structurally distinct (Galzi et al., 1991; Revah et al., 1991; White and Cohen, 1992; Jones and Westbrook, 1995; Baenziger and Chew, 1997; Auerbach and Akk, 1998; Williams and Akabas, 1999; Krampfl et al., 2000; Bianchi et al., 2001; Wilson and Karlin, 2001; Chang et al., 2002; Arevalo et al., 2005; Muroi et al., 2006; Purohit and Grosman, 2006). Several lines of evidence suggest that different gates occlude ion flow in the closed and desensitized states (Auerbach and Akk, 1998; Wilson and Karlin, 2001; Purohit and Grosman, 2006).
We used the accessibility of engineered cysteines in order to distinguish whether the GABAA receptors with picrotoxin trapped were in a closed or desensitized conformation. Previously, we showed that pCMBS− does not react with the M3 segment Cys mutant α1E303C in the closed state (Williams and Akabas, 1999). Furthermore, when pCMBS− is applied in the presence of GABA, the GABA currents during the application of pCMBS− are identical with GABA currents in the absence of pCMBS− (Fig. 4 B). However, the subsequent GABA-induced currents are markedly reduced compared with the GABA currents before pCMBS− application. Based on the absence of effect on the currents during the pCMBS− application, we inferred that pCMBS− must react in a desensitized state from which the channels rarely return to the open state. Thus, we infer that α1E303C is accessible to react in a desensitized state but not in the closed state. We showed in Fig. 4 E that in the picrotoxin-trapped state that pCMBS− does not react with α1E303C receptors. The reactivity, or lack thereof, of α1E303C in the picrotoxin-trapped state is consistent with it being a closed state where the cysteine is inaccessible to react with pCMBS− as opposed to an activated/desensitized state where the cysteine would have been reactive.
To further characterize the picrotoxin-trapped state we used another cysteine mutant, α1V280C (25'), in the M2 α helix that extends above the level of the membrane. We had previously shown that the rate of reaction of MTSET+ with this residue was seven times faster in the presence of GABA than in the absence of GABA (Bera et al., 2002). We exposed picrotoxin-trapped α1V280C receptors to an MTSET+ concentration for a duration of time that would have resulted in >80% reaction if the channels were in the activated state. After removal of MTSET+ and GABA-induced washout of the trapped picrotoxin, we found that the cysteine had not been modified (Fig. 5 E). In contrast, when we applied a higher concentration of MTSET+ that would react completely with picrotoxin-trapped α1V280C channels if they were in the closed state, we found that the cysteine was modified. Thus, the cysteine was accessible for modification, but in the picrotoxin-trapped channels it reacted at a rate similar to the rate at which it reacted with α1V280C channels in the closed state. It is important to recognize that during the time course of the sulfhydryl accessibility experiments the channels are undergoing transitions between multiple states. Thus, the reaction rates reflect the ensemble average of states present during the period of application. We cannot determine precisely in which state reaction is occurring. For α1V280C, the sevenfold difference in MTSET+ reaction rates in the absence and presence of GABA indicates that the ensemble average of states is different in the two conditions. The MTSET+ reactivity of α1V280C in the picrotoxin-trapped state is consistent with the inference that the picrotoxin-trapped channels are in a state resembling the closed state. Thus, we conclude that the gate that traps picrotoxin in the channel after the removal of GABA is a closed state gate. This gate must lie between the 2' level where picrotoxin binds and the extracellular end of the channel.
As outlined in the Introduction, inferences as to the location of the closed channel gate in the members of the Cys-loop receptors have varied depending on the channel studied and the approach used. Our current results for the GABAA receptor imply that a closed state gate exists between the 2' level and the extracellular end of the channel. Based on cysteine accessibility experiments, we had previously concluded that the ability of the sulfhydryl reagents to react with the 2' Cys mutant in the absence of GABA implied that the closed state gate was at a more cytoplasmic position than the 2' level (Xu and Akabas, 1996). There are several aspects of our previous work that limited the strength of the conclusion regarding the position of the gate. First, we had not measured the reaction rates of the sulfhydryl reagents and therefore did not know the extent to which the rates in the presence and in the absence of GABA differed. Second, we did not take the measures necessary to preclude the possibility that the reaction in the absence of GABA was truly in the closed state and not during either spontaneous openings or during brief “agonist-induced openings” where the sulfhydryl reagents could have acted as weak partial agonists. This would have entailed measuring the spontaneous open probability of the Cys mutant channels and applying the sulfhydryl reagents with a strong competitive antagonist to prevent the sulfhydryl reagents from activating the channels by binding in the agonist binding sites. In retrospect, the fact that pCMBS− reacted with α1V257C (2') in the absence of GABA but pCMB− did not, may have been a clue that pCMBS− could act as a weak agonist at least for the α1V257C channels (Xu and Akabas, 1996).
In the ACh receptor based on cysteine accessibility experiments similar conclusions were reached that the location of the gate was at a position more cytoplasmic than 2' (Akabas et al., 1994; Pascual and Karlin, 1998; Wilson and Karlin, 1998; Zhang and Karlin, 1998). Although rates were measured in some of these experiments they did not measure the spontaneous open probability of the Cys mutants nor did they perform the experiments in the presence of a competitive antagonist that would have prevented the reagents from acting as weak agonists. Based on the methanethiosulfonate ethylammonium (MTSEA+) reaction rate with the ACh receptor 2' position, αT244C, being significantly greater than 10 M−1s−1 in the absence of ACh and on differences in reaction rates due to application from the extracellular and cytoplasmic ends of the channel in the absence and presence of ACh, Wilson and Karlin (1998) concluded that the closed channel gate was more cytoplasmic than the 2' position. A potential weakness in these studies was that the inference that the extracellular side of the gate is below the 2' M2 position was largely based on the reactivity of the ACh receptor 2' cysteine substitution mutant, αT244C, with MTSEA+ in the absence of ACh (Wilson and Karlin, 1998). The rates of other MTS derivatives tested with this cysteine substitution mutant were significantly slower in the absence of ACh than in its presence. A doubly charged thiosulfonate, 2-aminoethyl-2-aminoethanethiosulfonate (AEAETS2+), reacted with αT244C at 1.25 M−1s−1 in the absence of ACh and 63,000 M−1s−1 in the presence of ACh (Pascual and Karlin, 1998). This might be consistent with the presence of a gate between the extracellular end of the channel and the 2' position. It is possible that MTSEA+, like other organic ammonium compounds, may act as a weak partial agonist of the ACh receptor. We previously observed that the quaternary ammonium MTS reagent MSTET+ displayed agonist activity with many of the ACh receptor M2 cysteine substitution mutants (Akabas et al., 1994). Perhaps little or no current was seen during the MTSEA+ applications to αT244C in the absence of ACh because the very fast open state reaction rate, 16800 M−1s−1, led to rapid reaction and complete inhibition of the channels as soon as they opened (Pascual and Karlin, 1998). These potential problems may limit the strength of the conclusion that in the ACh receptor there is no closed channel gate more extracellular than 2'. It should be noted that our present results do not rule out the existence of an additional, more cytoplasmic gate as was inferred from the ACh receptor SCAM experiments (Wilson and Karlin, 1998).
Using engineered histidines to create Zn2+ binding sites in ACh-5-HT3 chimeric channels it was concluded that the closed channel gate was more cytoplasmic than the 2' level (Paas et al., 2005). It is, however, clear from the experimental records shown in the published work that there is a significant level of spontaneous opening that can be seen when Zn2+ is applied and the holding currents decrease. Thus, it is likely that Zn2+ was reaching the engineered histidines during spontaneous openings, rather than in the closed state as was inferred in the paper. This undermines their conclusion that the gate is more cytoplasmic than 2' in the chimeric channel.
In the 5-HT3 receptor, in order to avoid the possibility of the MTS reagents acting as weak partial agonists and to reduce the potential for spontaneous channel opening, the closed state cysteine accessibility experiments were done in the presence of a high affinity, competitive antagonist (Panicker et al., 2002). These experiments were consistent with the extracellular side of the gate being at the 14' level, i.e., near the middle of the channel (Panicker et al., 2002).
Finally, structural evidence from the 4-Å resolution electron density maps indicates that the narrowest region in the closed ACh receptor is in the middle of the channel between the 9' and 14' residues (Miyazawa et al., 2003; Unwin, 2005). In this region the pore narrows to ∼6 Å in diameter. Several theoretical molecular dynamics studies have suggested that hydrated monovalent cations could not pass through a region this narrow at a significant rate, suggesting that it may form a hydrophobic gate (Amiri et al., 2005; Corry, 2006). Our present results, using a different functional approach, open channel blocker trapping, are consistent with a gate located in the middle of the channel as was inferred from the ACh receptor cryoelectron microscopic studies and from the 5-HT3A receptor SCAM experiments (Panicker et al., 2002; Miyazawa et al., 2003).
If picrotoxin can be trapped in the closed channel near the 2' level, it implies that in the closed state there must be a vestibule near 2' that can accommodate an ∼9-Å diameter picrotoxin molecule. This vestibule may not change size significantly during channel gating, consistent with the tight packing and low mobility that we previously inferred for this region (Goren et al., 2004) and that was suggested for the 5HT3 receptor (Panicker et al., 2004). Channel gating may involve movements of the M2 segments in a region more extracellular than the 6' residue.
In summary, we conclude that picrotoxin can be trapped in the closed GABAA receptor channel. To trap picrotoxin in its binding site near the cytoplasmic end of the channel, at the 2' level, there must be a gate located at a more extracellular position. Given that the reactivity of substituted cysteines is consistent with the picrotoxin-blocked channel being in the closed state, we suggest that this gate is a closed state gate. This is consistent with the 5-HT3A receptor SCAM experiments and with the 4-Å ACh receptor structure that inferred the existence of a gate in the middle of the channel in the 9'-14' region (Panicker et al., 2002; Miyazawa et al., 2003). It is likely that the gate location is similar in all members of the Cys-loop receptor superfamily. Establishing the location of the channel gate now focuses attention on where conformational changes must occur in order to transduce agonist binding in the extracellular domain into opening of the channel gate.
We thank Michaela Jansen, David Liebelt, Nicole McKinnon, David Reeves, and Paul Riegelhaupt for discussions and comments on this manuscript and Rachel Berkowitz for technical assistance.
This work was supported by National Institutes of Health grants NS30808 and GM77660 to M.H. Akabas.
Olaf S. Andersen served as editor.
Abbreviations used in this paper: ACh, acetylcholine; MTSEA+, methanethiosulfonate ethylammonium; MTSET+, methanethiosulfonate ethyltrimethylammonium; pCMBS−, p-chloromercuribenzenesulfonate; SCAM, substituted cysteine accessibility method.