Article Information

• PDF (156 kb)
• Full Text with Large Figures

PubMed

• Articles by Pei Tang
• Articles by Yan Xu

Related Articles

• Effects of Volatile Anesthetic on Channel Structure of Grami...

  Effects of Volatile Anesthetic on Channel Structure of Gramicidin A Biophysical Journal, Volume 83, Issue 3, 1 September 2002, Pages 1413-1420 Pei Tang, Pravat K. Mandal and Martha Zegarra Abstract Volatile anesthetic agent, 1-chloro-1,2,2-trifluorocyclobutane (F3), was found to alter gramicidin A channel function by enhancing Na transport (Tang et al. 1999. . 77:739–746). Whether this functional change is associated with structural alternation is evaluated by circular dichroism and nuclear magnetic resonance spectroscopy. The circular dichroism and nuclear magnetic resonance results indicate that at low millimolar concentrations, 1-chloro-1,2,2-trifluorocyclobutane causes minimal changes in gramicidin A channel structure in sodium dodecyl sulfate micelles. All hydrogen bonds between channel backbones are well maintained in the presence of 1-chloro-1,2,2-trifluorocyclobutane, and the channel structure is stable. The finding supports the notion that low affinity drugs such as volatile anesthetics and alcohols can cause significant changes in protein function without necessarily producing associated changes in protein structure. To understand the molecular mechanism of general anesthesia, it is important to recognize that in addition to structural changes, other protein properties, including dynamic characteristics of channel motions, may also be of functional significance. Abstract | Full Text | PDF (227 kb)

• Structural Consequences of Anesthetic and Nonimmobilizer Int...

  Structural Consequences of Anesthetic and Nonimmobilizer Interaction with Gramicidin A Channels Biophysical Journal, Volume 76, Issue 5, 1 May 1999, Pages 2346-2350 Pei Tang, Virgil Simplaceanu and Yan Xu Abstract Although interactions of general anesthetics with soluble proteins have been studied, the specific interactions with membrane bound-proteins that characterize general anesthesia are largely unknown. The structural modulations of anesthetic interactions with synaptic ion channels have not been elucidated. Using gramicidin A as a simplified model for transmembrane ion channels, we have recently demonstrated that a pair of structurally similar volatile anesthetic and nonimmobilizer, 1-chloro-1,2,2-trifluorocyclobutane (F3) and 1,2-dichlorohexafluorocyclobutane (F6), respectively, have distinctly different effects on the channel function. Using high-resolution NMR structural analysis, we show here that neither F3 nor F6 at pharmacologically relevant concentrations can significantly affect the secondary structure of the gramicidin A channel. Although both the anesthetic F3 and the nonimmobilizer F6 can perturb residues at the middle section of the channel deep inside the hydrophobic region in the sodium dodecyl sulfate micelles, only F3, but not F6, can significantly alter the chemical shifts of the tryptophan indole N-H protons near the channel entrances. The results are consistent with the notion that anesthetics cause functional change of the channel by interacting with the amphipathic domains at the peptide-lipid-water interface. Abstract | Full Text | PDF (143 kb)

• General Anesthetic Binding to Gramicidin A: The Structural R...

  General Anesthetic Binding to Gramicidin A: The Structural Requirements Biophysical Journal, Volume 78, Issue 4, 1 April 2000, Pages 1804-1809 Pei Tang, Roderic G. Eckenhoff and Yan Xu Abstract There is a distinct possibility that general anesthetics exert their action on the postsynaptic receptor channels. The structural requirements for anesthetic binding in transmembrane channels, however, are largely unknown. High-resolution H nuclear magnetic resonance and direct photoaffinity labeling were used in this study to characterize the volatile anesthetic binding sites in gramicidin A (gA) incorporated into sodium dodecyl sulfate (SDS) micelles and into dimyristoylphosphatidylcholine (DMPC) bilayers, respectively. To confirm that the structural arrangement of the peptide side chains can affect anesthetic binding, gA in nonchannel forms in methanol was also analyzed. The addition of volatile anesthetic halothane to gA in SDS with a channel conformation caused a concentration-dependent change in resonant frequencies of the indole amide protons of W9, W11, W13, and W15, with the most profound changes in W9. These frequency changes were observed only for gA carefully prepared to ensure a channel conformation and were absent for gA in methanol. For gA in DMPC bilayers, direct [C]halothane photolabeling and microsequencing demonstrated dominant labeling of W9, less labeling of W11 and W13, and no significant labeling of W15. In methanol, gA showed much less labeling of any residues. Inspection of the 3-D structure of gA suggests that the spatial arrangements of the tryptophan residues in the channel form of gA, combined with the amphiphilic regions of lipid, create a favorable anesthetic binding motif. Abstract | Full Text | PDF (237 kb)

• …more

Copyright © 1999 The Biophysical Society. All rights reserved.
Biophysical Journal, Volume 77, Issue 2, 739-746, 1 August 1999

doi:10.1016/S0006-3495(99)76928-6

Channels, Receptors, and Transporters

Distinctly Different Interactions of Anesthetic and Nonimmobilizer with Transmembrane Channel Peptides

Pei Tang^,  , Jian Hu, Serguei Liachenko and Yan Xu

Department of Anesthesiology and Critical Care Medicine, Department of Pharmacology, University of Pittsburgh, Pittsburgh, Pennsylvania 15261 USA

Address reprint requests to Dr. Pei Tang, W-1357 Biomedical Science Tower, University of Pittsburgh, Pittsburgh, PA 15261. Tel.: 412-383-9798; Fax: 412-648-9587.

The century-long quest for the mechanisms of general anesthesia remains a clinical and scientific challenge (Franks and Lieb, 1994). To date, the vast majority of investigations at the molecular level are indirect in nature, inferring mechanisms based on the observed effects of general anesthetics. Not enough attention has been paid to the direct interaction at the molecular level between an anesthetic molecule and a central nervous system site. In particular, when transmembrane proteins are under consideration, little is known about the nature of the sites of such interactions (Eckenhoff and Johansson, 1997).

Sensitivity to general anesthetics has been used as one of the important criteria for identifying proteins that are possibly involved in general anesthesia. A superfamily of ligand-gated ion channels (Franks and Lieb, 1996), which are highly sensitive to general anesthetics, has been the subject of intensive investigations. At the current stage, however, these ligand-gated channels are too complex to allow for structural analysis. A simplified membrane protein model may therefore prove useful in the study of the unifying nature of anesthetic-protein interaction that may be generalized for more complicated systems.

As a simplified model, the well-characterized gramicidin A channel (HCO-l-Val¹-Gly²-l-Ala³-d-Leu⁴-l-Ala⁵-dVal⁶-l-Val⁷-d-Val⁸-l-Trp⁹-d-Leu¹⁰-l-Trp¹¹-d-Leu¹²-l-Trp¹³d-Leu¹⁴-l-Trp¹⁵-NHCH₂CH₂OH) offers at least the following four advantages, which are unavailable in many other systems: 1) high achievable nuclear magnetic resonance (NMR) spectral resolution of the peptide permits absolute identification of individual amino acid residues along the transmembrane channel (Arseniev et al,Cross, 1994,Hinton, 1996), thus allowing direct quantitation of the site-specific interaction between anesthetics and the channel peptide; 2) a spectroscopically resolved protein-lipid interface allows testing of popular hypotheses relating to the protein-lipid interaction; 3) an environment that is unique to membrane proteins but not to soluble proteins can be unambiguously defined at the interfacial region, where protein, lipid, and water come into contact; and 4) because gramicidin forms a functional channel with resolved three-dimensional structures, it provides a unique system for determining the structural and functional consequences of anesthetic interaction.

The use of structurally similar pairs of general anesthetics and nonimmobilizers (originally referred to as “nonanesthetics”) has attracted much attention in recent years (Koblin et al,Tang et al,Ionescu et al,Raines, 1996,Forman and Raines, 1998,North and Cafiso, 1997). Although each anesthetic or nonimmobilizer cannot verify the relevance of a molecular site to general anesthesia, many nonimmobilizers as a group, in comparison with their anesthetic pairs, may help to identify the characteristics that are shared by all anesthetics but not by the nonimmobilizers (Xu et al). In protein-free lipid vesicles, we have shown that anesthetics and nonimmobilizers had different submacromolecular distribution in the lipid bilayers (Tang et al). The anesthetics, such as 1-chloro-1,2,2-trifluorocyclobutane (F3) and isoflurane, preferentially distribute to regions with easy access to the aqueous phase, whereas the structurally similar nonimmobilizers, such as 1,2-dichlorohexafluorocyclobutane (F6) and 2,3-dichlorooctofluorobutane (F8), dwell in the lipid core of the membrane. We have suggested that such a different distribution of anesthetics and nonimmobilizers may ultimately affect their affinity for other membrane constituents, particularly transmembrane proteins.

In the present study, we analyzed the specific interaction of F3 and F6 with gramicidin A, using intermolecular truncated driven nuclear Overhauser effects (TNOEs) (Xu and Tang, 1997). The functional consequences of sodium transport across the gramicidin A channel after interacting with F3 or F6 were also investigated, using the magnetization inversion transfer (MIT) experiments (Hinton et al).

l-α-Phosphatidylcholine (PC) and l-α-phosphatidylglycerol (PG) (egg-sodium salt) were purchased from Avanti Polar Lipids (Alabaster, AL). Gramicidin A was purchased from Calbiochem (La Jolla, CA). Other chemicals, including tripolyphosphate (PPP) and DyCl₃ (for preparation of ²³Na shift reagent), 2,2,2-trifluoroethanol (TFE), 2,2-dimethyl-2-silapentane-5-sulfonic acid (DSS), NaCl, and Na₂HPO₄ were obtained from Sigma (St. Louis, MO). D₂O was purchased from Cambridge Isotope Laboratories (Andover, MA). F3 and F6 (Fig. 1) were purchased from PCR (Gainesville, FL). All compounds were used without further purification.

Display large version of this figure

Figure 1

Chemical structures of the anesthetic 1-chloro-1,2,2-trifluorocyclobutane (F3) and the structurally similar nonimmobilizer 1,2-dichlorohexafluorocyclobutane (F6).

To obtain high-resolution ¹H-NMR spectra of gramicidin A for spectral identification of different residues, gramicidin A was incorporated into sodium dodecyl sulfate (SDS) micelles by the same procedure as described previously (Tang et al). The procedure ensures the formation of the head-to-head β6.3 dimers that are believed to be identical to the channel structure found in the lipid bilayers. For the TNOE experiments, sonicated vesicles of PC and PG in a 4:1 mole ratio, prepared as described previously (Xu and Tang, 1997), was used. The mean vesicle diameter was 130–160nm. To incorporate gramicidin A channel into the vesicles, aliquots of 50mM gramicidin solution in TFE were added, and the vesicle suspension was incubated at 50°C for 3h. Thereafter, the samples were cooled to room temperature and dialyzed in three exchanges against 4 liters of buffer for 6h. Complete removal of TFE was confirmed by ¹⁹F NMR. The final gramicidin concentration was 400μM in 60mM total lipid concentration, with pH adjusted to 6.5. For each NMR experiment, 2ml of vesicle suspension was placed in a gas-tight, 10-mm-diameter, high-precision NMR tube (Wilmad Glass Co., Buena, NJ), with ∼11.5ml vapor space above the suspension. F3 or F6 was added to a final concentration of 2mM in the vesicle suspension.

For MIT measurements, large unilamellar vesicles (LUVs) were prepared to ensure sufficient ²³Na signal from inside the vesicles. The extrusion method (Mayer et al) was employed using a LiposoFast extrusion device (Avestin, Ottawa, Cananda) with polycarbonate filters of defined pore sizes. The size and homogeneity of the LUV population were confirmed by light scattering, using a Coulter N4SD particle size analyzer (Coulter Electronics, Hialeah, FL). For a typical sample, the total lipid concentration was 66mM. The solutions inside and outside the vesicles contained 100mM NaCl and were buffered with 10mM K₂HPO₄/H₃PO₄ to a pH of 8.2.

A special procedure was followed to incorporate gramicidin A into LUV, so that no contamination exists from even a trace amount of organic solvents, which are often general anesthetics. Briefly, PC and PG in a 4:1 molar ratio were first mixed in chloroform and then dried to a thin film under a stream of N₂ gas. Residual chloroform was further removed by high vacuum. After hydration with a buffer with a precalculated Na⁺ concentration, aliquots of gramicidin A stock solution (25mM in TFE) were added to reach the desired gramicidin-to-lipid ratio. The mixture was then incubated at 50°C for 3h. Thereafter, the mixture was rapidly frozen in CO₂/acetone, lyophilized overnight at −50°C, and vacuumed at room temperature for an additional 24h. The completely dried samples were then rehydrated without shaking for 2–3h before the LUV extrusion. The final concentrations of gramicidin A ranged from 0 to 75μM.

The ²³Na chemical shift reagent, Na₇[DyPPP₂], was prepared as previously described (Gupta and Gupta, 1982,Chu et al) and used for the MIT experiments. The concentration of the shift reagent outside the vesicles was 3.9mM. For each MIT experiment, 0.6ml LUV suspension was used in a high-precision 5-mm NMR tube, which had 1.9-ml vapor space above the suspension. F3 and F6 were titrated into the LUV suspension to a total concentration in suspension of 6.9 and 6.3mM, respectively. Because the ratio of molar solubility of F3 in lipid and in water is ∼3000 (Tang et al), it can be estimated that for 6.9mM overall concentration of F3 in a 66mM lipid vesicle suspension, there are 27×10⁻⁶ moles of F3 per mole of water or, equivalently, 1.5mM F3 in the aqueous phase. This concentration is clinically relevant (Kendig et al).

The MIT and TNOE experiments were conducted at 25°C and 37°C, respectively, using an Otsuka-Chemagnetics (Fort Collins, CO) CMXW-400SLI spectrometer equipped with a four-nucleus probe (J S Research, Boston, MA). The resonance frequencies for ²³Na and ¹⁹F are 106.1 and 377.4MHz, respectively. To determine the ¹H resonance frequencies of gramicidin A in PC/PG vesicles, a sample with a 1:33 peptide-to-lipid ratio was prepared and used for magic angle spinning (MAS) ¹H NMR measurements. A selective excitation with a 2.25-ms, 820-mG square pulse was centered at 10ppm to increase the detection sensitivity of the indole and backbone amide protons and to minimize the water signal (Le Guerneve and Seigneuret, 1996).

The intermolecular ¹⁹F-{¹H} TNOE was measured between ¹⁹F of F3 or F6 and ¹H of gramicidin in selected frequency regions. In the TNOE pulse sequence, P_ts(¹H)-P_90°(¹⁹F)-Acq(¹⁹F)-T_R, a selected group of protons was saturated by a highly frequency-selective adiabatic pulse train, P_ts(¹H), which consisted of low-power, 90-ms, hyperbolic secant-shaped inversion pulses (Silver et al,Noggle and Schirmer, 1971) interleaved with 10-ms delays. The total saturation time, t_s, was varied among 11 predetermined values: t_s=0, 0.5, 1, 2, 3, 4, 5, 6, 8, 10, and 14s, the order of which was randomized in the experiments. After saturation, a 90° excitation pulse was applied to ¹⁹F, followed by ¹⁹F data acquisition. The nominal 90° pulse length for ¹⁹F was 25μs. For all ¹⁹F spectra, 16 scans were summed in 8192 complex points, with a recycle delay of T_R to ensure full relaxation of the ¹⁹F magnetization after each excitation. To minimize any unforeseen effects of T₁ relaxation, the T_R was varied such that t_s+T_R=constant=20s, which was ∼6–8 times the T₁ of ¹⁹F in the lipid vesicle suspension.

The MIT pulse sequence (Hinton et al) consisted of three nonselective 90° pulses, P(90)_x-t₁-P(90)_x-t₂-P(90)_x-Acq. In the presence of 3.9mM DyPPP₂ to separate the ²³Na signals from the inside and outside of the vesicles (²³Na_in and ²³Na_out, respectively), the transmitter frequency was placed on the resonance of ²³Na_out. The nominal 90° pulse was 11μs. The duration t₁ was precisely set to the reciprocal of twice the chemical shift difference between ²³Na_in and ²³Na_out. The mixing time t₂ was incremented from 0.015 to 500ms; ∼30 mixing times were used for each experiment. Typically, the repetition delay between two acquisitions was 400ms; the spectral width was 4kHz; and 512 complex points were collected and zero-filled once before Fourier transformation.

The method for processing the TNOE data has been elaborated previously (Xu and Tang, 1997). Briefly, for one I spin and N equivalent S spins (e.g., I=¹⁹F and S=¹H), the NOE, η, measured by the I spin while the S spins are saturated for a period of t_s, is given by

(1)

For the MIT experiments, one considers the modified Solomon equations for the magnetization inside (I) and outside (S) the vesicles:

(2)

(3)

If S magnetization is inverted, we have (assuming T₁≈T′₁, which is true for ²³Na_in and ²³Na_out)

(4)

(5)

Parameters from nonlinear regression are expressed as “the best estimates±standard errors of estimates” (Glantz and Slinker, 1990). To determine the effects of F3 and F6 on a given parameter, the t statistics are calculated as the ratio of the difference between the best estimates of that parameter to the standard error of the difference. A p value of <0.05 is considered statistically significant.

Specific interaction between anesthetic (or nonimmobilizer) molecules and gramicidin A can be quantified by intermolecular truncated driven NOE (Navon et al,Xu and Tang, 1997). Site selection can be achieved by narrow-band adiabatic saturation of a particular group of protons without affecting others. To demonstrate that such selections are feasible for gramicidin A, high-resolution ¹H spectra of gramicidin A in SDS micelles were used. Fig. 2 shows a stack plot of spectra in the regions of backbone and indole amide protons. The vertical arrows indicate the frequencies of the selective saturation by the hyperbolic secant-shaped pulse train. Clearly, with careful adjustment of pH to avoid base-catalyzed exchange with water (Arseniev et al,Hinton, 1996), saturation is well localized to the protons selected. Although high ¹H resolution is easily achievable in SDS, intermolecular NOE build-up is more favorable in a lipid bilayer environment. In the channel conformation, gramicidin A forms head-to-head β-6.3 helical dimers (Arseniev et al). Recent investigations in lipids (Weinstein et al,Killian et al,Cross, 1994,Cross, 1997,Ketchem et al,Mobashery et al) showed that the channel has the same secondary structure in lipid bilayers as in SDS micelles. Given the narrow ¹H chemical shift range, it can be assumed that the chemical shifts for most protons in the gramicidin A channel in SDS does not differ greatly from those in lipid vesicles. This is particularly true for the indole N-H protons, as we have observed a distinct indole N-H peak in high-resolution MAS ¹H NMR spectra of gramicidin channels in PC/PG lipid bilayers. As reported by others (Le Guerneve and Seigneuret, 1996), the peak is at the same chemical shift range as shown in Fig. 2 for the indole N-H protons and is absent in pure lipid sample without gramicidin. Using the same adiabatic pulse train for site selection in a series of intermolecular ¹⁹F-{¹H} TNOE experiments, we measured selective cross-relaxation rates between ¹⁹F of F3 or F6 and ¹H of gramicidin A in PC/PG vesicles (Fig. 3). In the absence of gramicidin A, selective saturation at 8.79 and 10.03ppm, which would correspond to resonance of the backbone and indole N-H protons, respectively, produced no NOE in either F3 or F6. After the addition of 400μM gramicidin A, the anesthetic F3 gave rise to a 〈∑σ〉 of 0.0105±0.0014s⁻¹ (n=8) and 0.0106±0.0007 (n=12) with backbone and indole amide protons, respectively. The nonimmobilizer F6, again, showed no measurable cross-relaxation with either the indole or the backbone amide protons. It should be pointed out that although the intermolecular NOE is small, as would be expected, the NOE in F3 is not due to artifacts such as nonspecific saturation through lipids or through intramolecular ¹⁹F-{¹H} NOE. The best proof of this is that the same experiments in the absence of gramicidin showed no observable changes in F3 intensity (open symbols in Fig. 3).

Display large version of this figure

Figure 2

High-resolution ¹H spectra of gramicidin A channel in SDS micelles, demonstrating the excellent selective saturation of individual proton groups by the hyperbolic secant-shaped pulse trains. Arrows above each spectrum indicate the frequency at which the selective saturation is centered. To emphasize the narrow selectivity, the frequency axis is plotted in units of Hz relative to the water resonance at 401.102MHz. Thus the frequency of W15 indole amide proton is at 10.10ppm, and that of A3 and A5 backbone amide proton is at 9.42ppm.

Display large version of this figure

Figure 3

Truncated driven ¹⁹F-{¹H} NOE selected in the backbone amide proton region at 8.79ppm (□, ■, ♦) and the indole N-H proton region at 10.03ppm (○, ●, ▴) to ¹⁹F in F3 (□, ■, ○, ●) or F6 (♦, ▴). Open and filled symbols are measurements in the absence and presence of gramicidin, respectively. With 400μM gramicidin A, 〈∑σ〉 between F3 and the backbone and indole amide protons was 0.0105±0.0014 (n=8) and 0.0106±0.0007 (n=12), respectively. Solid lines are best fit to the data using Eq. (1). The dashed line is a visual guide indicating no NOE build-up in either F3 in the absence of gramicidin or in F6 in the presence of gramicidin. Error bars show the standard error of the mean.

The distinction between F3 and F6 was also apparent in their ability to affect the function of the gramicidin A channel. Fig. 4 shows two stack plots of ²³Na NMR spectra of MIT experiments. In the absence of gramicidin A, the unfacilitated sodium transport was too slow on the NMR time scale to affect the ²³Na_in signal when the outside ²³Na_out was returning to the thermal equilibrium state after its inversion (right). In the presence of gramicidin A, however, the intensity profile of the ²³Na_in signal is modulated by the relaxation of ²³Na_out (left) due to transport of the inverted magnetization. Representative ²³Na_in signal profiles under the control condition and in the presence of F3 and F6 are plotted in Fig. 5. As shown, F3 at physiological concentrations produced a significant increase (p<0.05) in the unidirectional rates of ²³Na transport across the gramicidin A channels, whereas a similar concentration of F6 had little effect on the transport.

Display large version of this figure

Figure 4

Stack plots of ²³Na⁺ magnetization inversion transfer (MIT) spectra in the presence (left) and absence (right) of gramicidin channels. Intra- and extravesicular ²³Na⁺ signals are separated by 3.9mM DyPPP₂, a sodium shift reagent. The inset at the upper left corner shows the detailed changes in the intensity profile of ²³Na⁺_in due to Na⁺ transport.

Display large version of this figure

Figure 5

The signal intensity profiles of ²³Na_in⁺ are plotted as a function of ²³Na_out⁺ inversion recovery time, τ. Solid lines are best fit to the data using Eq. (4). At 75μM gramicidin A in 66mM PC/PG large unilamellar vesicles, 6.9mM F3 increased the ²³Na apparent influx rate constant, k_i, from 182.1±15.4 of control to 222.8±21.7s⁻¹ (n=3, p<0.05), and the apparent efflux rate constant, k_e, from 149.7±7.2 to 191.7±12.2s⁻¹ (n=3, p<0.001). A similar concentration of F6 (6.3mM) did not significantly affect the sodium transport, with k_i=176.9±17.8s⁻¹ (n=6, p>0.5) and k_e=162.2±8.9s⁻¹ (n=6, p>0.2). Error bars show the standard error of the mean.

Prodigious evidence suggests that general anesthetics exert their primary action by interacting with excitable membrane proteins (Franks and Lieb, 1994,Franks and Lieb, 1996,Mihic et al). The characteristics of such interactions, however, remain unidentified. Despite its structural simplicity and apparent lack of clinical relevance to general anesthesia, gramicidin A channel showed different functional responses to F3 and F6 in our MIT experiments. The parallelism between pharmacological profiles of these agents and their effects on sodium transport suggests that the characteristics of anesthetic-protein interaction found in this simplified system can potentially be generalized. For example, in lipid bilayers, it has been shown (Tang et al,Xu and Tang, 1997,North and Cafiso, 1997) that the anesthetics are in rapid exchange among all submacromolecular environments, with certain preference for the amphipathic lipid-water interface (Tang et al,Xu and Tang, 1997,Yoshino et al). Nonimmobilizers, in contrast, partition deep into the lipid core. This distinction may result from different distributions of partial atomic charges in the anesthetic and nonimmobilizer molecules. The structures in Fig. 1 illustrate a general tendency commonly found in anesthetics and nonimmobilizers, that is, general anesthetics often possess permanent or inducible displacement of partial charges, whereas nonimmobilizers are likely to be apolar (Trudell et al). Thus the four hydrogen atoms (more electropositive) and four halogen atoms (more electronegative) on two separate sides of F3 molecule make F3 more adaptable to the membrane interface, where a large dipolar potential exists. It has been suggested that modulation of the membrane dipole potential by anesthetics with polarizable or permanent dipole moments may play a significant role in the action of general anesthetics (Cafiso, 1998).

The difference in submacromolecular distribution of anesthetics and nonimmobilizers is likely to control their ability to interact with the transmembrane channel peptide. As we have shown previously with the channel surface map, there are four amphipathic domains at each end of the gramicidin A channel (Xu et al). These domains are attributable mainly to four tryptophan residues in each gramicidin molecule. The π-electron distributions in the indole rings may interact with the electropositive moiety of the anesthetic molecules by mechanisms similar to the cation-π type of interaction (Cubero et al). Indeed, our TNOE results in this study (Fig. 3) show that the anesthetic F3 interacts specifically with the tryptophan side chains, whereas the nonimmobilizer F6 showed no measurable interaction in this region. In separate 2D NOESY experiments with gramicidin A in SDS (Tang et al), we have found the same tendency under high-resolution conditions. The NOESY spectra revealed that whereas both F3 and F6 caused changes in chemical shift of the backbone N-H resonance of Val⁷ deep inside the bilayer, only F3, but not F6, produced significant changes in the interfacial tryptophan indole N-H proton resonance.

Interaction with tryptophan residues in the gramicidin channel may prove to be nontrivial. The critical roles of tryptophan in the gramicidin channel function are at least twofold. First, each of the indole rings is uniquely oriented, with its N-H bond directed toward the bilayer surface (Hu et al). Such orientation favors hydrogen bonding to the hydrophilic interface, presumably to water molecules that either are at the membrane surface or penetrate into the amphiphilic domain along the lipid-peptide interface. Second, the side chains of the tryptophan residues are extended parallel to the bilayer surface and are likely to be hydrogen-bonded to the lipid molecules (Meulendijks et al,Scarlata, 1991). These extended tryptophan side chains anchor the channel in the bilayer and orient the channel with respect to the surface. The tryptophan hydrogen bonding and indole electric dipole moments have been shown (Hu and Cross, 1995) to stabilize cations in the binding sites near the channel entrance and substantially reduce the potential energy barrier at the bilayer center. It has also been shown that the dynamics of the indole rings cause significant fluctuations in the energy of stabilization at the binding site, resulting in a possible mechanism for the rapid control of the channel conductance (Hu and Cross, 1995). Thus, based on our TNOE results (Fig. 3), it is conceivable that because of their amphipathic properties and their preferential distribution to the interfacial region, anesthetics, but not nonimmobilizers, can target the tryptophan side chains and alter their functional association with the peptide-lipid-water interface.

Our MIT experiments of sodium transport across gramicidin channel provide direct evidence in support of this argument. The transport rates determined by the MIT method are consistent with the simple diffusion model described by the Goldman-Hodgkin-Katz equation. For 10mM intra- and extravesicular [K⁺], 100mM intravesicular [Na⁺], and 127mM extravesicular [Na⁺] (because of the shift reagent), the zero-current potential can be estimated to be 4.7mV, provided that the permeability ratio of K⁺ to Na⁺ in gramicidin A channels is ∼3.5 (Myers and Haydon, 1972,Urban et al). At this potential, Ussing’s flux-ratio criterion (Ussing, 1949) predicts that the k_i-to-k_e ratio should be 1.2. Our results under the control condition are in excellent agreement with this prediction (Fig. 5, k_i/k_e≈182/150=1.21).

The rate constants k_i and k_e measured in the MIT experiments can also be used to estimate the absolute reaction rate constant for a single-ion transport across a single gramicidin A channel. Goulian et al measured the dimer lifetime (τ) of gramicidin A and [Gly¹]gramicidin A, as well as the dimer formation rate (f) of [Gly¹]gramicidin A. Although the lifetimes of gramicidin and [Gly¹]gramicidin are different, the formation rate constant of the two should be similar (Dimer formation rate constant=f/N_m², where N_m is the number of monomers). Using a lipid surface density of 0.7nm² per lipid dimer (Huang et al) and the gramicidin-to-lipid ratio used by Goulian et al, we estimated a dimer formation rate constant of 3×10⁻⁴mol⁻¹s⁻¹, based on the measured formation rate of f≈0.4mol/s. At steady state, fτ equals the number of gramicidin dimers (Goulian et al). Therefore, for τ≈8–10s (Goulian et al), ∼20–25% of gramicidin molecules (i.e., ∼16μM monomers in our case) form ion-conducting channels at any given time (i.e., ∼8μM functional channel dimers). Thus, for a rate constant of ∼200s⁻¹ measured by our MIT experiments, the absolute reaction rate constant through a single channel can be estimated to be k_i^abs=k_i[Na⁺]_e/[channel]=3.2×10⁶s⁻¹. According to the absolute reaction rate theory, k_i^abs=κ(kT/h)exp(−ΔG^≠/RT), where ΔG^≠ is the free energy of activation associated with the transport, and κ is the probability that a sodium ion will advance through the channel, rather than exiting from the same side of the channel, after reaching the top of the ΔG^≠ barrier. Because the gramicidin A channel acts as a symmetrical electrical image barrier (Hinton et al), it follows that κ≈0.5. Thus the rate constants measured by the MIT experiments correspond to a ΔG^≠ of ∼34kJ/mol, which is within the range reported in the literature (Hinton et al).

Microscopically, sodium transport is believed to involve five distinct steps (Andersen, 1983,Buster et al): 1) diffusion through the aqueous phase to the channel entrance, 2) association with the channel entrance (binding), 3) translocation through the channel (in a single file), 4) dissociation from the channel, and 5) diffusion away from the channel. Patch-clamp measurements, with sensitivity high enough to monitor single-channel events, suggest that the transport is diffusion-limited (Andersen, 1983), with the aqueous convergence conductance being the most significant factor in determining the permeability characteristics of the gramicidin A channel. Thus one possible interpretation of the anesthetic and nonimmobilizer effects on transport is that anesthetics, but not nonimmobilizers, can facilitate tryptophan (particularly indole ring) interaction with the interfacial water to stabilize the sodium binding at the channel entrance, thereby increasing the unidirectional transport rates (Fig. 5). A similar increase in ion transport was also found for a wide range of general anesthetics, including halothane, enflurane, methoxyflurane, and ethanol (Hunt and Veiro, 1986).

Infrared spectroscopic studies (Buchet et al) have shown that volatile anesthetics can disrupt the hydrogen bonds at the center of the lipid membrane, where the carbonyls are involved in the head-to-head dimerization of the gramicidin channel. These results have been taken to support an earlier finding (Bradley et al) that anesthetics at physiological concentrations can reduce the lifetime of N-acetyl-gramicidin dimers by a factor of 2 without affecting the single-channel current. This reduction was interpreted as anesthetic suppression of ion transport across the gramicidin channel, an apparent contradiction of the transport results found in this study with F3 and in other studies with a number of volatile anesthetics (Hunt and Veiro, 1986). The discrepancy may result from the difference between gramicidin and N-acetyl-gramicidin; the latter is known to form unstable channels (Urry, 1971). However, it has been shown (Finkelstein and Andersen, 1981) that a gramicidin channel is capable of transporting 2.6×10⁶ions/s, with a time constant of 0.38μs for a single file of ions. In contrast, the lifetime of gramicidin dimers is ∼10s (Goulian et al). Thus, on the time scale of ion translocation (step 3), the gramicidin channel can be viewed as a stable (or fixed) structure (Andersen, 1983). Therefore, a reduction by a factor of 2 in channel lifetime by anesthetics (Bradley et al) is unlikely to be the mechanism for anesthetic effects on channel conductance.

In conclusion, a pair of structurally similar anesthetic and nonimmobilizer interacts differently with a transmembrane channel peptide. Only the anesthetic, but not the nonimmobilizer, can interact specifically with the amphipathic domains near the channel-lipid-water interface. This interaction has direct functional consequences.