Nonanesthetic volatile compounds fail to inhibit movement in response to noxious stimulation at concentrations predicted to induce anesthesia from their oil-water partitioning. Thus they represent tools to determine whether molecular models behave like the targets that mediate in vivo anesthetic actions. The effects of volatile anesthetics and nonanesthetics were examined in two experimental models in which anesthetics interact directly with proteins: the pore of the nicotinic acetylcholine receptor and human serum albumin.
Wild-type mouse muscle nicotinic receptors and receptors containing pore mutations (alphaS252I + betaT263I) were studied electrophysiologically in membrane patches from Xenopus oocytes. Patch currents evoked by brief pulses of acetylcholine were measured in the presence of enflurane and two nonanesthetics, 1,2-dichlorohexafluorocyclobutane and 2,3-dichlorooctafluorobutane. Nonanesthetic interactions with human serum album were assessed by quenching of intrinsic protein fluorescence.
Both anesthetic and nonanesthetic volatile compounds inhibited wild-type and alphaS252I + betaT263I mutant nicotinic channels but displayed different selectivity for open versus resting receptor states. Median inhibitory concentrations (IC50s) in wild-type nicotinic receptors were 870+/-20 microM for enflurane, 37+/-3 microM for 1,2-dichlorohexafluorocylcobutane, and 11.3+/-5.6 microM for 2,3-dichlorooctafluorobutane. For all three drugs, ratios of wild-type IC50s to mutant IC50mut ranged from 7-10, and ratios of wild-type IC50s to predicted anesthetic median effective concentrations (EC50s) ranged from 1.8-2.3. 1,2-Dichlorohexafluorocyclobutane quenched human serum albumin with an apparent dissociation constant (Kd) of 160+/-11 microM. The ratios of dissociation constants to predicted EC50s for the nonanesthetics were within a factor of two of the dissociation constant:EC50 ratios calculated for halothane and chloroform from previous published results.
In two models in which anesthetics bind to protein sites, both anesthetic and nonanesthetic volatile drugs cause similar steady state effects with potencies that are predicted by hydrophobicity. These protein sites do not sterically discriminate between anesthetic and nonanesthetic drugs. However, differential state-selective actions on ion channel targets may underlie the distinct in vivo effects of anesthetics and nonanesthetics.
NEARLY a century ago, Meyer and Overton noted the strong correlation between general anesthetic potency and the oil - water partition coefficient. The Meyer - Overton correlation led to the commonly accepted idea that these drugs act at a hydrophobic site in neuronal tissues. On a molecular scale, hydrophobic sites where general anesthetics might act include the lipids that form neuronal membranes and the intrinsic proteins that function in the membrane. In model lipid membranes, anesthetics alter physical properties such as microviscosity (measured as a spectroscopic order parameter), but the mechanism(s) that presumably translate membrane perturbations into altered neuronal activity remain unknown. Many neuronal membrane proteins are sensitive to anesthetics in the clinically relevant concentration range, but demonstration of functional protein sites where anesthetics bind directly has proved challenging. Among the membrane proteins that are sensitive to anesthetics, only one such site has been identified, in the channel of the nicotinic acetylcholine receptor (nAChR). 
Compounds that disobey the Meyer - Overton correlation are highly hydrophobic and predicted to be potent anesthetics, but they do not prevent animals from moving in response to stimulation. Examples of these socalled nonanesthetics or nonimmobilizers include long-chain alkanes and alcohols (the “cut-off” effect) and a recently described group of halogenated volatile molecules. [5–8]Why do these compounds disobey the Meyer - Overton correlation? This question is critical to understanding the molecular mechanism of anesthesia, because a comprehensive mechanistic theory must explain both anesthetic and nonanesthetic actions. Thus the nonanesthetics represent important tools for testing the validity of molecular models of in vivo sites where general anesthetics act.
Molecular theories based on a lipid site for anesthetics suggest that nonanesthetics should differ from anesthetics in their ability to perturb lipid bilayers. Supporting these theories are data showing that both alcohols beyond the “cut-off”[6,9]and volatile nonanesthetics fail to disorder model lipid membranes. Theories of anesthesia based on direct protein site interactions require that these sites somehow discriminate between nonanesthetics and anesthetics. Nonanesthetics and anesthetics have been shown to differ in their modulation of some ligand-gated ion channels, [10,11]but studies of anesthetic and nonanesthetic volatile drug effects in well-established protein site models are lacking.
In this study, we compare the behavior of anesthetic and nonanesthetic volatile compounds using two experimental models where anesthetics interact directly with protein sites. The first model is the peripheral nAChR, an intrinsic membrane protein that forms a transmembrane cation channel when gated by bound acetylcholine. The nAChR is a useful model to study anesthetic actions because its structure and kinetic mechanisms are well established, as are the roles of several structural domains in its function (Figure 1). The nAChR structural motif is shared by a superfamily of anesthetic-sensitive ligand-gated ion channels, including gamma-aminobutyric acidA, glycine, 5-HT3, and neuronal nicotinic receptors. 
Anesthetics have been shown to interact directly with the nAChR transmembrane pore, inhibiting the flow of cations during gating. The presence of a discrete protein site where anesthetics bind in the pore was first suggested by pharmacologic and kinetic studies [14,15]and confirmed by mutagenesis experiments that altered specific amino acids in the M2 hydrophobic domains that form the pore lining. When hydroxylated amino acids (serine = S, threonine = T) near the middle of the M2 domains on alpha ([Greek small letter alpha] S252) and beta ([Greek small letter beta] T263) subunits (Figure 1[right]) are mutated to hydrophobic isoleucines (I), anesthetics bind about 10 times more tightly, as demonstrated in rapid-perfusion electrophysiology experiments. Despite their remarkable sensitivity to hydrophobic blockers, the [Greek small letter alpha] S252I +[Greek small letter beta] T263I mutant receptors display essentially normal gating and desensitization. [4,16]
The second protein site model we used is human serum albumin (hSA), a lipid-free soluble protein. Anesthetic interactions with a site on hSA can be assessed by quantifying anesthetic quenching of fluorescence emitted from the single tryptophan in each molecule. The fluorescence spectrum of this tryptophan is also sensitive to the protein folding pattern of hSA. Thus the possibility that anesthetics and nonanesthetics significantly alter protein folding was also tested.
Materials and Methods
Nicotinic Receptor Experiments
Receptor sensitivities to a volatile anesthetic and two volatile nonanesthetics were assessed using voltage-clamped membrane patches from Xenopus oocytes that expressed the four subunits of either wild-type or mutated mouse muscle nAChRs. A perfusion system that can produce submillisecond changes in the drug concentrations was used to elicit currents from patches expressing receptors.
cDNAs and Site-directed Mutagenesis. cDNAs encoding [Greek small letter alpha], [Greek small letter beta], [Greek small letter gamma], and [Greek small letter delta] subunits of the mouse muscle nAChR were obtained from Dr. James Boulter (Salk Institute, San Diego, CA) and subcloned into a pSP64 plasmid containing Xenopus globin noncoding sequences (pSP64T ) to enhance translation (provided by Dr. J. McLaughlin, Tufts University Medical School, Boston, MA). The [Greek small letter alpha] S252I mutant cDNA was made by oligonucleotide-directed mutagenesis in pBluescript (Stratagene, La Jolla, CA), as described by Tomaselli et al., and a 800 bp Bsu361-Bgl II cDNA fragment containing the mutated M2 coding sequence was transferred into the wild-type [Greek small letter alpha]-pSP64T construct. The [Greek small letter beta] T263I mutant cDNA was provided by Dr. Cesar Labarca (California Institute of Technology, Pasadena, CA) in pGEM2-SP6.
Xenopus Oocyte Expression. Wild-type and mutant nAChRs were expressed on the surface of Xenopus oocytes after injection of oocytes with messenger RNA (mRNA) mixtures encoding the four nAChR subunits. Detailed methods were previously described. Adult hcg-primed female Xenopus laevis frogs (Xenopus One, Ann Arbor, MI) were anesthetized in ice-cold 0.2% aminobenzoic acid ethyl ester, and several lobes of ovary were removed through a small abdominal incision. All procedures were approved by the Massachusetts General Hospital Animal Care Committee. Oocytes were enzymatically treated (with 1 mg/ml collagenase D for 1 - 2 h) to remove connective tissues, and stage V and VI oocytes were selected for injection. Messenger RNAs were transcribed in vitro from linearized cDNA templates using SP6 RNA polymerase (Ambion, Austin, TX). Subunit mRNAs were isolated using affinity beads (BIO-101, Vista, CA) mixed stoichiometrically at 2 [Greek small letter alpha]:[Greek small letter beta]:[Greek small letter gamma]:[Greek small letter delta] and microinjected into oocytes (25 - 50 nl). After incubation for 48 - 96 h, oocytes were stripped manually of their vitelline membranes and used for electrophysiologic recordings.
Patch-clamp Electrophysiologic Recordings. Electrophysiologic recordings were made at room temperature (20–22 [degree sign]C). Borosilicate patch pipettes were polished to give open tip resistance of 2–5 M Omega. Oocyte membrane patches were pulled in the outside-out configuration and held at -50 mV. Inside and outside buffers were symmetrical K-100 (97 mM KCl, 1 mM MgCl2, 0.2 mM EGTA, 5 mM K-HEPES, pH 7.5). Currents through the patch-clamp amplifier (Axopatch-200A; Axon Instruments, Foster City, CA) were filtered (8-pole bessel, 2 kHz) and digitized at 5 - 10 kHz using a 586-class personal computer, a 12-bit A/D converter (National Instruments, Austin, TX), and custom software.
Rapid Perfusion. Rapid acetylcholine and anesthetic concentration jumps at the experimental patch surface were achieved using a computer-activated, piezo-driven theta tube. Patches were perfused continously with control solution (K-100 with or without volatile drug) through one lumen of the theta tube. A computer signal actuated the piezo to rapidly position the other theta tube lumen (acetylcholine in K-100 with or without volatile drug) before the patch. Perfusate exchange times (open pipette junction current method) were 0.2 to 0.5 ms. Nicotinic receptors were activated with high concentrations of acetylcholine (200 [micro sign]M) to avoid interference from changes in apparent acetylcholine agonist site affinity (KACh) induced by volatile drugs or by mutations. [4,21]The acetylcholine exposure period was 300 ms, and patches were “recovered” in control perfusate for 5 - 15 s between acetylcholine exposures. Volatile drug effects were assessed with drugs present both in control and acetylcholine perfusates.
Data Analysis and Statistics
Each patch studied under a given set of acetylcholine/volatile drug conditions was exposed to these drugs sequentially 8 - 16 times with a recovery period in between each exposure. The ensemble of current traces was aligned at the midpoints of the rapid current rise (channel opening) and averaged. Control ensemble average currents (acetylcholine alone) were assessed before and after experiments where patches were exposed to volatile drugs. Data were not analyzed if the two peak control ensemble average currents differed by more than 10%. For concentration - response studies, data from at least three patches from different oocytes were pooled and averaged for each volatile drug concentration studied.
Depending on the number of kinetic components in the current decay, either single (Equation 1) or double (Equation 2) exponential functions were fitted to the decay portion of current data. where Ipeakis the peak current and [Greek small letter tau]desis the time constant of desensitization. and Equation 3where Ifastand Isloware the current amplitudes, and [Greek small letter tau]fastand [Greek small letter tau]sloware the time constants of the two kinetic decay components. L sub [infinity] is the steady state current after desensitization.
Inhibitory drug concentration-response data were analyzed by fitting logistic functions (Equation 4) to control-normalized data:
Linear and nonlinear least-squares fitting and Student's t test analyses were performed using Origin (Microcal Inc., Northampton, MA) software on a 586-class personal computer. All results are reported as means +/- SD.
Enflurane (CHF2OCF2CHFCI) was purchased from Anaquest (Murray Hill, NJ). The nonanesthetics 1,2-dichlorohexafluorocyclobutane (F6) and 2,3-dichlorooctafluorobutane (F8) were purchased from PCR (Gainesville, FL). Saturated solutions of volatile drugs in K-100 buffer were prepared by stirring in glass-stoppered flasks overnight. Saturated stocks were diluted with K-100 immediately before use, mixed by inversion three times, and immediately transferred to a glass reservoir syringe to minimize time for evaporation. The perfusion system was entirely of glass and Teflon to minimize adsorption, and both solutions flowed continuously through the theta tube. Gas chromatography (Hewlett Packard 5890 with a J and W DB-WAX122 7033 column, Palo Alto, CA) was used to assay volatile compound concentrations, as described by Raines. 
Estimation of Volatile Drug Losses from Aqueous Solutions
Because the nonanesthetics have very small buffer-gas partition coefficients, their use in aqueous solutions represents an unusual challenge. Gas chromatography showed that we had <15% losses of F6 and F8 during our experiments (which typically lasted about 3 min) where mixing and handling of nonanesthetic solutions were minimized. Because of the minuscule volume of membrane patches (approximately 1 x 10-191), no significant depletion of drug from the flowing perfusates occurred.
Human Serum Albumin Experiments
Lyophilized fatty-acid free hSA was purchased from Sigma Chemical Company (St. Louis, MO). Aliquots of saturated nonanesthetic solutions in phosphate-buffered saline (pH 7.4) were diluted with phosphate-buffered saline and concentrated hSA to a final concentration of 5 [micro sign]M hSA. The mixtures were transferred immediately to a 10-mm pathlength quartz cell that was sealed with a Teflon stopper. Care was taken to completely eliminate any air space above the solution. Intrinsic fluorescence emission spectra were obtained using a Fluoro-Max 2 spectrofluorimeter (Jobin Yvon-Spex, Edison, NJ). The sample compartment was thermostatically maintained at 22 +/- 1 [degree sign]C. Fluorescence was excited at 295 nm to selectively excite the tryptophan residue, and emission spectra were obtained between 305 - 450 nm. For quantitative determinations of nonanesthetic quenching of hSA fluorescence, the fluorescence was recorded at the emission maximum of 338 nm using a slit width of 2 nm and an integration time of 2 s. For control experiments using L-tryptophan, an emission wavelength of 355 nm was used.
Both Volatile Anesthetics and Nonanesthetics Inhibit Wild-type nAChR Channels
In the absence of volatile drugs, rapid acetylcholine perfusion of membrane patches expressing wild-type nAChRs elicited currents that reached their peak within several milliseconds and then underwent a monoexponential decay as a result of receptor desensitization (Figure 2, control curves). Nearly all of the receptors open in response to 200-[micro sign]M acetylcholine. Peak control currents (Ipeakcntl, Table 1) varied from patch to patch because of varying numbers of receptors in each patch. Desensitization time constants ([Greek small letter tau]des, Table 1) and the extent of desensitization also varied somewhat. From pooled results on 24 patches, the average desensitization rate (1/[Greek small letter tau](des)) for wild-type control currents in these experiments was 8 +/- 1.2 s (-1). Desensitization rates represent maxima because we used saturating acetylcholine concentrations.
In the presence of 780 [micro sign]M (1.4% vol/vol) enflurane, the shape of the acetylcholine-induced current trace was nearly identical to that of the control trace (Figure 2, enflurane), except that the overall current was scaled down and appears more noisy. The peak current in the presence of 780 [micro sign]M enflurane (1peakenf) was 55% of the peak control current in the example shown. Current decay in the presence of enflurane remained monoexponential at a rate the same as that in the absence of enflurane (8 +/- 1.5 s-1, n = 4). The observation that peak current was inhibited to the same extent as later currents suggests that (1) enflurane inhibits wild-type nAChRs before channels open or at least as fast as we can switch perfusates (>5,000 s-1), and (2) enflurane has little effect on the maximal agonist-induced desensitization rate.
Like enflurane, F6 and F8 both inhibited wild-type nAChRs (Figure 1), but the shape of the inhibited currents showed two exponential decay phases instead of one. At the F6 and F8 concentrations shown, modest inhibition of peak nAChR currents was observed (Ipeakvol/Ipeak(cntl), Table 1), but this inhibition rapidly deepened within 15 - 30 ms before further decaying at a slower rate. The slower current decay rates (1/[Greek small letter tau]slow) were near the control current desensitization rate (8 +/- 1.3 s-1and 7 +/- 1.0 s-1) and therefore were very likely a result of this process. The fast-current decay represents the onset of channel inhibition after acetylcholine activation, because volatile drugs were present before and during acetylcholine perfusion. Thus these nonanesthetic volatile drugs act as selective inhibitors of the open channel state.
After the fast inhibition phase was complete, the fraction of uninhibited current, relative to control, reached steady state. In some wild-type currents, desensitization was incomplete, so we directly compared steady state current estimates (I sub [infinity] in Equation 2) in the presence and absence of volatile compounds. This analysis yielded I sub [infinity]F6/Isub [infinity]cntl= 0.52 +/- 0.04 and I sub [infinity]F8/Isub [infinity]cntl= 0.27 +/- 0.05 for the currents in Figure 2. An alternative analysis was used for patches where desensitization was nearly or fully complete. Because the volatile drugs had negligible effects on desensitization time constants, we estimated the steady state inhibited current components by subtracting the fast decay components (Figure 2, dotted lines). The fractional steady state current was calculated as (Ipeakvolatile- Ifast)/Ipeakcntl, and yielded values close to those from the I sub [infinity] ratios shown in Figure 2(Table 1).
Both Volatile Anesthetics and Nonanesthetics Inhibit [Greek small letter alpha] S252I +[Greek small letter beta] T263I Mutant nAChR Channels
Control currents elicited by 200 [micro sign]M acetylcholine in patches expressing [Greek small letter alpha] S252I +[Greek small letter beta] T263I nAChRs showed features very similar to those from wild-type patches (Figure 3, controls). Currents increased rapidly to a peak and then monoexponentially decayed due to desensitization. The major difference noted was faster desensitization in [Greek small letter alpha] S252I +[Greek small letter beta] T263I nAChRs, with an average rate of 12 +/- 2.5 s-1(n = 18). Desensitization of [Greek small letter alpha] S252I +[Greek small letter beta] T263I nAChRs was consistently complete, resulting in I sub [infinity] values of zero.
Enflurane inhibited [Greek small letter alpha] S252I +[Greek small letter beta] T263I nAChRs in a similar manner to that seen in wild-type receptors (Figure 3), but inhibitor was observed at much lower enflurane concentrations. Peak currents were reduced and monoexponential current decay was observed at rates close to those seen in the absence of enflurane (13 +/- 1.8 s-1at 47 [micro sign]M enflurane, n = 4).
The [Greek small letter alpha] S252I +[Greek small letter beta] T263I nAChRs were also more sensitive to the volatile nonanesthetics, F6 and F8. Furthermore, the nonanesthetics caused biexponential current decay in [Greek small letter alpha] S252I +[Greek small letter beta] T263I nAChRs, indicating that their open-state selective inhibition was maintained. In the presence of F6, the two current decay phases were readily separated at concentrations that partially inhibited the mutant receptors. Thus we could determine, as for wild-type receptors, that desensitization rates were not significantly changed in the presence of F6 (Table 2). In mutant receptor currents in the presence of F8, the separation of two distinct decay phases required a modification of the analysis strategy used for F6. At high concentrations of F8, apparent current decay time constants were decreased from control, but current was fully inhibited in this concentration range, making observation of slow decay impossible. At lower F8 concentrations that partially inhibited mutant receptors, the apparent decay time constants were only slightly lower than the control time constants. We therefore assumed that the slow current decay rate (1/[Greek small letter tau]slow) in the presence of F8 was the same as the control desensitization rate in the patch being studied (as established for both enflurane and F6), and we fixed this value in the two-exponential fit analysis.
Concentration Dependence of Steady State nAChR Inhibition by Volatile Drugs
In both wild-type and [Greek small letter alpha] S252I +[Greek small letter beta] T263I nAChRs, rapid-ensemble average currents like those in Figure 2and Figure 3were recorded over a range of volatile drug concentrations. For patches exposed to enflurane, steady state inhibition was derived by normalizing peak currents in the presence of drug to control peak currents in the same patch. For patches exposed to F6 and F8, steady state inhibition was assessed by subtracting the fast decay component before normalizing to control peak. Pooled data from at least three patches at each volatile drug concentration were averaged (Figure 4). Data were fitted with a logistic equation (Equation 4in Materials and Methods) and fitted parameters are reported in Table 3.
Neither F6 nor F8 fully inhibited wild-type nAChRs, even when present at saturated aqueous concentrations (Figure 4). Saturated F6 (240 [micro sign]M) inhibited 92 +/- 3.3%(n = 6) of current, whereas saturated F8 (50 [micro sign]M) inhibited 84 +/- 7.5%(n = 4) of current. The lack of full inhibitory efficacy might be due to the limited aqueous solubility of the nonanesthetic compounds (i.e., higher aqueous concentrations would cause full inhibition) or it could reflect a true lack of inhibitory efficacy. This question was addressed by using the [Greek small letter alpha] S252I +[Greek small letter beta] T263I mutant nAChR, which is more sensitive to channel inhibition by hydrophobic drugs. Both F6 and F8 fully (>or= to 98%) inhibited currents through the mutant receptors at or below their saturating concentrations.
The IC50mutfor enflurane in [Greek small letter alpha] S252I +[Greek small letter beta] T263I patches was 10 times lower than the wild-type median inhibitory concentration (IC50;Figure 4, Table 3). The mutant-to-wild-type sensitivity ratio determined for F6 was also 10. The estimated IC50mutfor F8 was only seven times lower than the wild-type IC50, but the uncertainty in this ratio was high. This uncertainty arose from the poorly separated fast and slow decay phases in mutant ensemble average currents in the presence of very low F8 concentrations. The mutant-to-wild-type sensitivity ratio for F8 is not significantly different from those for enflurane and F6 (P > 0.4 by the Student's t test).
Concentration Dependence of Fast Current Decay Rates in the Presence of F6 and F8
Both the rates and amplitudes of the fast-current decay phases (1/[Greek small letter tau]fastand Ifast, Equation 2) increased as F6 and F8 concentrations increased. In both wild-type and mutant experiments, plots of 1/[Greek small letter tau]fastversus [F6] or [F8] were linear (Figure 5), suggesting a rate-limiting, bimolecular interaction between the drugs and their site of action:Equation 5where D is the volatile drug, s is the channel receptor site, and DS is the inhibited channel. The observed time constant for onset of inhibition ([Greek small letter tau]fast) in this simple scheme depends on the sum of binding and unbinding processes:Equation 6
Thus the slope of a line fitted to data in Figure 5estimates the apparent on-rate (kon), whereas the y-axis intercept represents the apparent off-rate (koff). In wild-type receptors, linear least-squares analysis reveals apparent on-rates for F6 and F8 of 2.4 (+/- 0.3) x 106M-1[middle dot] s-1and 1.3 (+/- 0.2) x 106M-1[middle dot] s-1, respectively. The apparent on-rates for F6 and F8 in mutant receptors were close to those seen in wild-type receptors. In contrast, the apparent off-rates for F6 and F8 in mutant receptors were about three times lower than those in wild-type receptors.
Quenching of Human Serum Albumin Intrinsic Fluorescence by Nonanesthetics
The binding of the nonanesthetics, F6 and F8, to the soluble protein hSA was examined by assessing their ability to quench the intrinsic fluorescence of the protein's single tryptophan moiety. Equilibration with F6 reduced the fluorescence emission of hSA in a dose-dependent manner (Figure 6A). Figure 6B shows the difference spectrum obtained by subtracting the spectrum of hSA in the presence of 120 [micro sign]M F6 from a control spectrum without F6. The two spectra have the same maxima at 338 nm and nearly identical shapes. At the highest F6 concentration studied (213 [micro sign]M), hSA intrinsic fluorescence was quenched 60 +/- 2%. We determined the affinity of F6 for the fluorescence quenching site on hSA by fitting the quenching data in Figure 7to the following equation, from Johansson et al. :Equation 7where F is the fluorescence intensity, Qmaxis the maximum fraction of fluorescence that can be quenched, and Kdis the apparent dissociation constant of F6 for the site. With an unrestrained nonlinear least-squares fit, we obtained a Kdof 160 +/- 11 [micro sign]M and a Qmaxof 1.05 +/- 0.038.
To rule of the possibility that quenching of the hSA tryptophan was independent of F6 protein binding, we evaluated the ability of F6 to quench the fluorescence of free L-tryptophan. Figure 7shows that in contrast to its affect on hSA intrinsic fluorescence, F6 did not quench the fluorescence of free L-tryptophan.
F8 also quenched hSA intrinsic fluorescence without causing any spectral shift (data not shown). Quenching of hSA fluorescence increased with [F8], but at the highest experimental F8 concentration we could study (40 [micro sign]M), only 27 +/- 1.8% quenching was observed. Because the the F8 quenching data lacked a plateau, we could not confidently determine both Q (max) and Kdwhen fitting with Equation 7. By assuming a Qmaxof 1.0 (see Discussion), we estimated a maximum Kdfor F8 of 110 [micro sign]M.
In both the nAChR and hSA experiments, our data to not support a model where protein sites discriminate sterically between nonanesthetics and anesthetics. Here we discuss the detailed interpretation of each set up results individually and then return to the broader issue of the molecular mechanisms of anesthesia.
Interpretation of nAChR experiments
We have shown that both volatile anesthetics and nonanesthetics inhibits currents through nAChR channels. At first it appears that nonanesthetics have a lower inhibitory efficacy, because neither F6 nor F8 fully inhibit wild-type receptors at saturating aqueous concentrations. However, no plateau in inhibition effect is observed at high nonanesthetic concentrations (Figure 4). Furthermore, both F6 and F8 fully inhibit the [Greek small letter alpha] S252I +[Greek small letter beta] T2631 mutant nAChR channels, demonstrating that at a sufficiently sensitive site, their efficacy is no lower than that of volatile anesthetics. Thus the lack of full inhibition of wild-type nAChRs by nonanesthetics is due to their limited aqueous solubility (i.e., the saturated concentrations are less than the 20 x IC50needed to cause 95% inhibition).
A single site in the nAChR pore appears to be involved in steady state inhibition by anesthetic alcohols, volatile anesthetics, and volatile nonanesthetics. The ratios of wild-type IC50s to [Greek small letter alpha] S252I +[Greek small letter beta] T263I IC50mutare all about 10 times for enflurane, F6, and F8 (Table 3). Inhibition by long-chain alcohols in these two nAChRs results in the same sensitivity ratio, and isoflurane sensitivity is affected similarly by mutations at these sites in the nAChR pore. The simplest explanation for the observed mutational effects and the kinetic data are that all of these hydrophobic drugs interact directly with the M2 domain amino acid side chains that form the pore lining.
Although steady stage inhibition by both anesthetics and nonanesthetics is apparently mediated by the same site, the kinetic mechanisms of nAChR inhibition by anesthetics and nonanesthetics differ. Immediately after channels open, F6 and F8 appear to be quite weak inhibitors of nAChRs, whereas full steady state inhibition develops within several to tens of milliseconds. This is because the nonanesthetics preferentially inhibit the open state of nAChR, suggesting that the pore site is more accessible to them after the channel opens. A small degree of closed state inhibition is evident in the reduced peak currents (Figure 2, Figure 3), but most of the steady state inhibition develops after acetylcholine binding and channel opening. Similar open state-selective block has been observed for octanol. 
Unlike F6 and F8, enflurane shows no selectivity for inhibiting the open state. No fast decay component is observed, even at the lowest enflurane concentrations (20 [micro sign]M) used, whereas with the volatile nonanesthetics, selective open-channel inhibition onset is readily observed at 10–50 [micro sign]M (Figure 2, Figure 3). One possible explanation for enflurane's lack of apparent open state selectivity is that its on-rate may be much faster than konfor other drugs, but this is unlikely because the apparent konfor its isomer, isoflurane, is comparable to those measured for alcohols and nonanesthetics. We conclude that enflurane inhibits both resting (closed) and open nAChR states at comparable concentrations. Isoflurane also lacks significant open state selectivity. Inhibition of two nAChR state may explain why in enflurane concentration-response studies (Figure 4), Hill coefficients >1.0 are observed (Table 3). F6 or F8, which apparently inhibit a single kinetic state, result in Hill coefficients near 1.0.
The observation that the [Greek small letter alpha] S252I +[Greek small letter beta] T263I mutation causes the same sensitivity change (from wild type) for enflurane as it does for F6 and F8 suggest that the same site in the channel is involved in both closed and open state inhibition. How can two similar groups of molecules bind at the same site, yet have different kinetic mechanisms? The anesthetics may inhibit closed channels via a separate mechanism, such as desensitization, or they may somehow gain access to the channel site in the closed state. Supporting this hypothesis are data showing that nonanesthetics are significantly less potent than anesthetics as closed state desensitizers of nAChR. 
The apparent rates of current inhibition by F6 and F8 are well fit by a bimolecular interaction model in which binding is diffusion limited and unbinding is determined by the strength of the interactions between the inhibitor and its site. The apparent on-rates for F6 and F8 are similar to each other and in the same range reported for octanol (5.7 x 106M-1[middle dot] s-1), propofol (6 x 106M-1[middle dot] s-1) and isoflurane (2 x 106M-1[middle dot] s-1) block of nAChR channels. [4,15,20]Apparent on-rates for F6 and F8 are not significantly affected by channel mutations, whereas the apparent off-rates are significantly slower in the sensitive [Greek small letter alpha] S252I +[Greek small letter beta] T263I receptors compared with those in wild-type receptors. This observation is consistent with the idea that the volatile drugs are binding more tightly to the more hydrophobic mutant site. It should be noted that the uncertainties in our off-rate values derived from current decay analyses are high. This may explain why koff/kondoes not closely match IC50in our results.
Our data indicate that nAChR maximal desensitization rates are not significantly altered in the presence of volatile inhibitors. Other studies have shown that anesthetics enhance nAChR desensitization rates at low agonist concentrations, but this observation is not incompatible with the present results, which were obtained with saturating agonist concentrations. Isoflurane also changes desensitization rates modestly at high acetylcholine concentrations. Thus, although desensitization rates may be accelerated by anesthetics at low agonist occupancy, the maximal desensitization rate is nearly unchanged.
The wild-type IC50s for the volatile drugs are proportional to their Meyer - Overton predicted median effective concentrations (EC50s;Table 3, last column). Plotting log(IC50) against log(predicted EC (50))(Figure 8) produces a straight-line relation with slope that is indistinguishable from 1.0, the familiar form of the Meyer - Overton correlation. There is a less than twofold difference in IC50:EC50ratios for enflurane compared with the nonanesthetics. This difference is much smaller than the experimental differences in anesthetic potency observed in animals (nonanesthetics are at least five or six times less potent than predicted) and may be partially explained by loss of the very insoluble non-anesthetics from solution during the experiments. Another factor that may contribute to the different IC50:EC50ratios is that enflurane inhibits both closed and open nAChR states nearly equally, whereas the nonanesthetics preferentially inhibit the open state.
Interpretation of Human Serum Albumin Experiments
Although hSA is not an excitable membrane protein, it represents a protein site model that is devoid of lipid. Both F6 and F8 quench hSA fluorescence from its single tryptophan, as previously reported for halothane and chloroform. Because the nonanesthetics do not quench free tryptophan fluorescence in aqueous solutions, their quenching of hSA fluorescence cannot be attributed to simple surface effects or absorption of photons in the bulk solution. Thus quenching of hSA fluorescence by volatile anesthetics and nonanesthetics is due to direct protein interactions. Because nonanesthetics induce no spectral changes in the hSA fluorescence spectrum, we further conclude that F6 and F8 do not cause significant changes in hSA folding.
The value of Qmax[almost equal to] 1.0 (Figure 7) shows that F6 can fully quench hSA fluorescence. Thus F6 has the same quenching efficacy as halothane and chloroform, in agreement with other reports of its indole fluorescence-quenching properties. F8 has also been shown to be an efficient quencher of indole fluorescence in non-aqueous solvent. 
The Kdfor F6 quenching of hSA fluorescence is 160 +/- 11 [micro sign]M, which is about 10 times its predicted anesthetic EC50. Assuming that the Qmaxfor F8 is 1.0 (based on its ability to quench indole fluorescence in nonaqueous solvents), we calculate a maximum estimate for its fluorescence quenching Kdof 110 [micro sign]M, which is 24 times its anesthetic EC50. These ratios for the nonanesthetic drugs are comparable to the Kd:EC50ratios calculated for chloroform and halothane from published data (Table 4). F6 has a Kd:EC50ratio between those of chloroform and halothane, whereas the Kd:EC50ratio for F8 is close to that for halothane. The overall fourfold range in these ratios is higher than the ranges of IC50:EC50ratios in nAChR experiments (Table 3), suggesting that factors other than hydrophobicity may play a role in binding or quenching of hSA by volatile drugs. Nonetheless, the overall correlation with hydrophobicity is quite good, as shown by the log(Kd) versus log(EC50) plot with a slope near 1.0 for the four drugs (Figure 8, hSA data). This result suggests that hydrophobicity maintains a dominant role in determining volatile drug binding to hSA. Furthermore, despite the spread in Kd:EC50ratios shown in Table 4, there is clearly no consistent differentiation between anesthetics and nonanesthetics evident from these results.
Molecular Mechanisms of Anesthesia
During the last decade, the idea that anesthetics act in vivo by binding directly to protein sites on target molecules (such as ligand-gated ion channels) has gained in popularity. We find that at two protein sites where anesthetics are known to bind, nonanesthetic and volatile anesthetics cause similar effects on binding, with relative potencies that are predicted by the Meyer-Overton correlation. In contrast, recent studies on lipid bilayer models show that anesthetics and nonanesthetics differ in their ability to disorder acyl chains. Thus compared with our protein site models, lipid models seem to better reflect the in vivo differences in anesthetic and nonanesthetic actions. Nonetheless, direct protein interactions in vivo may underlie important actions of volatile anesthetic, and our results suggest some intriguing alternate interpretations.
Protein Sites Do Not Sterically Discriminate between Anesthetics and Nonanesthetics
In both models we studied, the steady state effects of nonanesthetic volatile compounds correlated with their predicted Meyer - Overton EC50s (Figure 8). Because both the nAChR IC50s and hSA Kds represent steady-state binding of drugs to these respective sites, we conclude that neither protein site distinguishes between anesthetic and nonanesthetic structures. Indeed, our data suggest that hydrophobicity is the major determinant of binding between volatile drugs and these protein sites. We know of no published reports in which protein-binding sites have been shown to be capable of sterically distinguishing between anesthetic and nonanesthetic volatile compounds.
In other ion channel experiments, the functional effects of anesthetics and nonanesthetics are distinguished, but it is not known whether these effects are mediated by direct drug binding to protein sites. For example, anesthetics enhance Torpedo nAChR desensitization kinetics (at low agonist concentrations), whereas nonanesthetics do not. In gamma-aminobutyric acidAreceptors, anesthetics enhance channel gating by low agonist concentrations, and nonanesthetics do not. [10,25]
Protein States May Discriminate between Anesthetics and Nonanesthetics
Although our interpretation of binding site affinities for anesthetics and nonanesthetics with nAChR is based on steady state current inhibition, the different state-dependent mechanisms observed in our rapid perfusion experiments suggests how their in vivo physiologic actions might differ. Suppose a central nervous system ion channel with a site similar to the nAChR channel site is an important target for anesthetics in vivo. If, like many synaptic channels, this target channel opens for only a few milliseconds before closing (or desensitizing), then slow open-channel selective inhibitors such as F6 and F8 will not have sufficient time to equilibrate with their binding site, which is accessible only during the transient open state. In contrast, anesthetic drugs such as enflurane inhibit the channel before it opens, enabling them to inhibit currents on the physiologically important time scale. This alternative interpretation of our data shows the value of using submillisecond drug superfusion switching to observe both transient states and state transitions in proteins with complex functional kinetics.
Protein Sites May Be Involved in Anesthetic Actions Other than Immobilization
Our results suggest that protein sites may not be involved in the immobilizing action of anesthetics (i.e., those measured by MAC assays). Nonetheless, direct protein binding may mediate important actions of general anesthetics. An obvious example is that binding at the muscle nAChR channel site explains the clinical observation that volatile anesthetics potentiate neuromuscular blockade by relaxants. Furthermore, Eger et al. recently proposed that the immobilizing actions of anesthetics and their amnestic actions occur at distinct sites in the nervous system. Nonanesthetics fail to immobilize animals, but there is evidence that they suppress memory formation in accordance with the Meyer - Overton rule. Thus the protein sites on nAChR and hSA may represent good models for putative protein sites where anesthetics cause amnesia, rather than those that cause immobility.
Lipids Can Discriminate between Anesthetics and Nonanesthetics
In contrast to our studies on protein models, recent studies in lipid bilayers seem to reflect the inactivities of nonanesthetics in vivo. Anesthetics disorder membrane lipids, whereas nonanesthetics fail to do so. Both theoretical model calculations and NMR studies suggest that anesthetics and nonanesthetics occupy different regions of the membrane bilayer, causing different pertubations in overall microviscosity. [28–30]These studies suggest that anesthetics concentrate at the aqueous - membrane interface, disrupting well-ordered headgroups and decreasing-order parameters for probes deep in the acyl-chain region of the bilayer. In contrast, nonanesthetics concentrate mainly in the less-ordered deep regions of the bilayer, where they have little effect on membrane properties. Perhaps occupation of these different depth domains of the membrane underlies the differential access to the nAChR channel in the closed state observed in this study.
It is important to note that microviscosity alone (or other physical properties of lipids) cannot be linked to anesthesia or even altered membrane protein function in most cases, because microviscosity changes caused by temperature do not have the same effects on animals or membrane proteins as those produced by anesthetics. The mechanism of anesthesia may be explained by forces in the bilayer or lipid - protein interactions that vary with the depth of the perturbant molecule. [31–33]If such “specific” lipid models of anesthesia are to be critically tested, we will need new methods to examine intermolecular forces in membranes.
The authors thank Carol Gelb and Nanditha Krishnan for technical assistance and Dr. Keith Baker for his comments on the manuscript.