Basic Anesthesia 2

Anesthetic Effects on Second Messenger–Activated Ion Channels

Ion channels can be activated by ligands present in the cytoplasm as well as by ligands present in the extracellular space. The intracellular ligands that activate these channels are generally chemical second messengers, including cyclic nucleotides, Ca²⁺ or H⁺ ions, inositol phosphates, and ATP. The structure of second messenger–activated ion channels is not as well understood as that of the voltage- or ligand-activated channels, and there is little information about anesthetic effects on these channels.

A potassium-selective channel (referred to as I_K(an)), found in snail neurons, that has many of the properties of a second messenger–activated channel is activated by clinical concentrations of volatile anesthetics.^121,133,134 I_K(an) shares many biophysical properties with a second messenger– activated potassium channel found in Aplysia neurons that is referred to as the S channel. Recent work by Yost and colleagues has shown that the S channel is also activated by clinical concentrations of volatile anesthetics.¹³⁵ The importance of volatile anesthetic activation of second messenger–activated potassium channels in invertebrates has now become apparent with the discovery of a large family of so-called “background potassium channels” in mammals. These mammalian potassium channels have a unique structure with two pore-forming domains in tandem plus four transmembrane segments (2P/4TM; Fig. 6C).¹³⁶ TOK1, a member of this family, was first shown by Yost and colleagues to be activated by volatile anesthetics.¹³⁷ The laboratory of Michel Lazdunski has studied the effects of a variety of volatile anesthetics on several members of the 2P/4TM family. They found that TREK-1 channels were activated by clinical concentrations of chloroform, diethyl ether, halothane, and isoflurane (Fig. 6B). In contrast, closely related TRAAK channels were insensitive to all the volatile anesthetics, and TASK channels were activated by halothane and isoflurane, inhibited by diethyl ether, and unaffected by chloroform. These authors went on to show that the C-terminal regions of TASK and TREK-1 contained amino acids essential for anesthetic actions on TASK and TREK-1 channels.¹³⁸ More recently, TREK-1 but not TASK was found to be activated by clinical concentrations of the gaseous anesthetics—xenon, nitrous oxide, and cyclopropane.¹³⁹ Thus, activation of background K⁺ channels in mammalian vertebrates could be an important and general mechanism through which inhalational anesthetics regulate neuronal resting membrane potential and thereby excitability; this effect could plausibly be a significant contributor to some components of the anesthetic state.

FIGURE 6. Volatile anesthetics activate background K+ channels. (Panel A) Halothane reversibly hyperpolarizes a pacemaker neuron from Lymnaea stagnalis (the pond snail) by activating IKan. (Panel B) Halothane (300 /µM) activates human recombinant TREK-1 channels expressed in COS cells. The figure shows current-voltage relationships with reversal potential (Vrev) of -88 mV, indicative of a K+ channel. (Panel C) Predicted structure of a typical subunit of the mammalian background K+ channels. Note the four transmembrane spanning segments (in black) and the two pore-forming domains (P1 and P2). Some but not all of these 2P/4TM K+ channels are activated by volatile anesthetics. (Panel D) Phylogenetic tree for the 2P/4TM family. (Reproduced with permission from Franks NP, Lieb WR. Background K+ channels: An important target for anesthetics? Nature Neurosci. 1999;2:395.)

One type of second messenger–activated channel, the calcium-dependent potassium channels, has been shown to be inhibited by clinical concentrations of anesthetics.¹⁴⁰ These large conductance potassium channels open in response to increases in cytoplasmic Ca²⁺ concentration and are important in modulating the shape and frequency of action potentials in the central nervous system. While a wide variety of anesthetics inhibit channel opening, this would tend to excite neurons and is thus unlikely to be important in the depressant effects of anesthetics. Anesthetic effects on these channels may contribute to the excitatory effects of low concentrations of anesthetics and to the convulsant properties of some anesthetic agents. Several other potassium-selective ion channels are also activated by second messengers, including ATP-activated channels and channels activated by muscarinic acetylcholine receptors, but the effects of anesthetics on these channels has not been delineated.

Summary

Second messenger–activated ion channels are a plausible target for anesthetic action. Recent evidence suggests that members of the 2P/4TM family of background potassium channels may be important in producing some components of the anesthetic state.

WHAT IS THE CHEMICAL NATURE OF ANESTHETIC TARGET SITES?

The Meyer-Overton Rule

Almost 100 years ago, Meyer and Overton independently observed that the potency of gases as anesthetics was strongly correlated with their solubility in olive oil (Fig. 7).^141,142 This observation has significantly influenced thinking about anesthetic mechanisms in two ways. First, because a wide variety of structurally unrelated compounds obey the Meyer-Overton rule, it has been reasoned that all anesthetics are likely to act at the same molecular site. This idea is referred to as the Unitary Theory of Anesthesia. Second, it has been argued that since solubility in a specific solvent strongly correlates with anesthetic potency, the solvent showing the strongest correlation between anesthetic solubility and potency is likely to most closely mimic the chemical and physical properties of the anesthetic target site in the CNS. Based on this reasoning, the anesthetic target site was assumed to be hydrophobic in nature

FIGURE 7. The Meyer-Overton rule. There is a linear relationship (on a log-log scale) between the oil/gas partition coefficient and the anesthetic potency (MAC) of a number of gases. The correlation between lipid solubility and MAC extends over a 70,000-fold difference in anesthetic potency. (Reproduced with permission from Tanfiuji Y, Eger EI, Terrell RC. Some characteristics of an exceptionally potent inhaled anesthetic: thiomethoxyflurane. Anesth Analg. 1977;56:387.)

The Meyer-Overton correlation suffers from two limitations: (1) it only applies to gases and volatile liquids because olive oil/gas partition coefficients cannot be determined for liquid anesthetics; (2) olive oil is a poorly characterized mixture of oils. To circumvent these limitations, attempts have been made to correlate anesthetic potency with water/solvent partition coefficients. To date, the octanol/water partition coefficient best correlates with anesthetic potency. This correlation holds for a variety of classes of anesthetics and spans a 10,000-fold range of anesthetic potencies.¹⁴³ The properties of the solvent octanol suggest that the anesthetic site is likely to be amphipathic, having both polar and nonpolar characteristics.

Exceptions to the Meyer-Overton Rule

Halogenated compounds exist that are structurally similar to the inhaled anesthetics yet are convulsants rather than anesthetics.¹⁴⁴ There are also convulsant barbiturates¹⁴⁵ and neurosteroids.¹⁴⁶ One convulsant compound, fluorothyl (hexafluorodiethylether), has been shown to cause seizures in 50% of mice at 0.12 vol%, but to produce anesthesia at higher concentrations (EC₅₀ = 1.22 vol%).¹⁴⁷ The concentration of fluorothyl required to produce anesthesia is approximately predicted by the Meyer-Overton rule. In contrast, several polyhalogenated alkanes have been identified that are convulsants but that do not produce anesthesia. Based on the olive oil/gas partition coefficients of these compounds, anesthesia should have been achieved within the range of concentrations studied.¹⁴⁸ The end point used to determine the anesthetic effect of these compounds was movement in response to a noxious stimulus (MAC). Interestingly, some of these polyhalogenated compounds do produce amnesia in animals.¹⁴⁹ These compounds are thus referred to as nonimmobilizers rather than as nonanesthetics. Several polyhalogenated alkanes have also been identified that anesthetize mice, but only at concentrations 10 times those predicted by their oil/gas partition coefficients;¹⁴⁸ these compounds are referred to as transitional compounds. The nonimmobilizers and transitional compounds have been proposed as a “litmus test” for the relevance of anesthetic effects observed in vitro to those observed in the whole animal.

In several homologous series of anesthetics, anesthetic potency increases with increasing chain length until a certain critical chain length is reached. Beyond this critical chain length, compounds are unable to produce anesthesia, even at the highest attainable concentrations. In the series of n-alkanols, for example, anesthetic potency increases from methanol through dodecanol; all longer alkanols are unable to produce anesthesia.¹⁵⁰ This phenomenon is referred to as the cutoff effect. Cutoff effects have been described for several homologous series of anesthetics including n-alkanes, n-alkanols, cycloalkanemethanols,¹⁵¹ and perfluoroalkanes.¹⁵² While the anesthetic potency in each of these homologous series of anesthetics shows a cutoff, a corresponding cutoff in octanol/water or oil/gas partition coefficients has not been demonstrated. Therefore, compounds above the cutoff represent a deviation from the Meyer-Overton rule.

A final deviation from the Meyer-Overton rule is the observation that enantiomers of anesthetics differ in their potency as anesthetics. Enantiomers (mirror-image compounds) are a class of stereoisomers that have identical physical properties, including identical solubility in solvents such as octanol or olive oil. Animal studies with the enantiomers of barbiturate anesthetics,^{153 154} ketamine,⁹⁴ neurosteroids,¹⁰³ etomidate,¹⁵⁵ and isoflurane¹⁵⁶ all show enantioselective differences in anesthetic potency. These differences in potency range in magnitude from a >10-fold difference between the enantiomers of etomidate or the neurosteroids to a 60% difference between the enantiomers of isoflurane. It is argued that a major difference in anesthetic potency between a pair of enantiomers could only be explained by a protein binding site (see Protein Theories of Anesthesia); this appears to be the case for etomidate and the neurosteroids. Enantiomeric pairs of anesthetics have also been used to study anesthetic actions on ion channels. It is argued that if an anesthetic effect on an ion channel contributes to the anesthetic state, the effect on the ion channel should show the same enantioselectivity as is observed in whole animal anesthetic potency. Early studies showed that the (+)-isomer of isoflurane is 1.5 to 2 times more potent than the (-)-isomer in eliciting an anesthetic-activated potassium current, in potentiating GABA_A currents, and in inhibiting the current mediated by a neuronal nicotinic acetylcholine receptor.¹⁰⁵,¹²¹ In contrast, the stereoisomers of isoflurane are equipotent in their effects on a voltage-activated potassium current and in their effects on lipid phase-transition temperature.¹²¹ Studies with the neurosteroids¹⁰³ and etomidate¹⁵⁵ show that these anesthetics exert enantioselective effects on GABA_A currents that parallel the enantioselective effects observed for anesthetic potency.

The exceptions to the Meyer-Overton rule do not obviate the importance of the rule. They do, however, indicate that the properties of a solvent such as octanol describe some, but not all, of the properties of an anesthetic binding site. Compounds that deviate from the Meyer-Overton rule suggest that anesthetic target site(s) are also defined by other properties including size and shape.

In defining the molecular target(s) of anesthetic molecules one must be able to account both for the Meyer-Overton rule and for the well-defined exceptions to this rule. It has sometimes been suggested that a correct molecular mechanism of anesthesia should also be able to account for pressure reversal. Pressure reversal is a phenomenon whereby the concentration of a given anesthetic needed to produce anesthesia is greatly increased if the anesthetic is administered to an animal under hyperbaric conditions. The idea that pressure reversal is a useful tool for elucidating mechanisms of anesthesia is based on the assumption that pressure reverses the specific physicochemical actions of the anesthetic that are responsible for producing anesthesia; that is to say, pressure and anesthetics act on the same molecular targets. However, recent evidence suggests that pressure reverses anesthesia by producing excitation that physiologically counteracts anesthetic depression, rather than by acting as an anesthetic antagonist at the anesthetic site of action.¹⁵⁷ Therefore, in the following discussion of molecular targets of anesthesia, pressure reversal will not be further discussed.

Lipid vs. Protein Targets

Anesthetics might interact with several possible molecular targets to produce their effects on the function of ion channels and other proteins. Anesthetics might dissolve in the lipid bilayer, causing physicochemical changes in membrane structure that alter the ability of embedded membrane proteins to undergo conformational changes important for their function. Alternatively, anesthetics could bind directly to proteins (either ion channel proteins or modulatory proteins), thus either (1) interfering with binding of a ligand (e.g., a neurotransmitter, a substrate, a second messenger molecule) or (2) altering the ability of the protein to undergo conformational changes important for its function. The following section summarizes the arguments for and against lipid theories and protein theories of anesthesia.

Lipid Theories of Anesthesia

The elucidation of the Meyer-Overton rule suggested that anesthetics interact with a hydrophobic target. To investigators in the early part of the twentieth century, the most logical hydrophobic target was a lipid. In its simplest incarnation, the lipid theory of anesthesia postulates that anesthetics dissolve in the lipid bilayers of biological membranes and produce anesthesia when they reach a critical concentration in the membrane. Consistent with this hypothesis, the membrane/gas partition coefficients of anesthetic gases in pure lipid bilayers correlate strongly with anesthetic potency.¹⁵⁸ This simple theory can account for anesthetics that obey the Meyer-Overton rule, but cannot account for anesthetics that deviate from this rule. For example, the cutoff effect cannot be explained by this theory because compounds above the cutoff can achieve membrane concentrations equal to those of compounds below the cutoff.¹⁵⁹ Similarly, enantioselectivity cannot be explained by this theory. Most importantly, this simplest version of the lipid theory does not explain how the presence of the anesthetic in the membrane is translated into an effect on the function of the embedded proteins.

Membrane Perturbation

More sophisticated versions of the lipid theory require that the anesthetic molecules dissolved in the lipid bilayer cause a change or perturbation in one or more physical properties of the membrane. According to this theory, anesthesia is a function of both the concentration of anesthetic in the membrane and the effectiveness of that anesthetic as a perturbant. This potentially could explain deviations from the Meyer-Overton rule, because nonanesthetics could achieve high concentrations in the membrane, but might not be effective perturbants. In examining this theory it is important to define explicitly the perturbation caused by an anesthetic. One can then test the relevance of a specific perturbation to the mechanism of anesthesia by measuring the perturbation caused by various compounds (anesthetics and nonanesthetics) and correlating perturbation with anesthetic potency. The specific perturbations of membrane structure that have been proposed to be causally related to the anesthetic state are briefly explored in the following section.

Membrane Expansion

Anesthetics dissolved in membranes do increase membrane volume. This occurs both because the anesthetic molecules occupy space and, in principle, because they produce changes in lipid packing and/or protein folding. The critical volume hypothesis is an attempt to correlate changes in membrane volume with anesthesia. This hypothesis predicts that anesthesia occurs when anesthetic dissolved in the membrane produces a critical change in membrane volume. Changes in membrane volume could compress ion channels and thus alter their function. Alternatively, increases in membrane thickness could alter neuronal excitability by changing the potential gradient across the plasma membrane.¹⁶⁰ Several studies have shown that anesthetics can produce changes in membrane volume.¹⁶¹ However, the amount of expansion caused by clinical concentrations of anesthetics is probably very small. One study of erythrocyte membranes showed that halothane (0.27 mM = 1.0 MAC) expanded the membranes by only 0.1%.¹⁶² Another study of erythrocyte membranes showed that both anesthetics and nonanesthetics (long-chain n-alkanols above the anesthetic cutoff) produced similar degrees of membrane expansion.¹⁶³ While clinical concentrations of anesthetics clearly produce membrane expansion, the small magnitude of anesthetic-induced membrane expansion, coupled with the inability of this theory to account for the cutoff effect, makes it unlikely that membrane expansion is the correct mechanism of anesthesia. A recent study by Cantor revisits this topic.¹⁶⁴ Based on thermodynamic modeling, he argues that anesthetics in biologic membranes preferentially distribute to the interface between lipid and aqueous phases. This distribution results in increased lateral pressure, which could alter the function of membrane-embedded ion channels. His calculations also suggest that nonimmobilizers should not show the same interfacial distribution. There is some experimental evidence showing that anesthetics, but not nonimmobilizers, do preferentially distribute to the lipid/aqueous interface in a membrane.¹⁶⁵ The relationship between these recent observations and anesthetic effects on protein function remains to be determined.

Membrane Disordering

Studies using nuclear magnetic resonance (NMR) spectroscopy¹⁶⁶ and electron spin resonance (ESR) spectroscopy¹⁶⁷ have shown that a variety of anesthetics can disorder the packing of phospholipids in lipid bilayers and in biological membranes. The decrease in membrane order (often referred to as an increase in membrane fluidity) can, in principle, alter the function of ion channels embedded in the lipid bilayer. The ability of anesthetics to fluidize lipid bilayers does show a modest correlation with anesthetic potency.¹⁶⁸ Membrane disordering can also account for the cutoff effect. Studies on synaptic membranes have shown that anesthetic alkanols (octanol, decanol, dodecanol) fluidize membranes, whereas nonanesthetic alkanols have either no effect on fluidity (tetradecanol) or a rigidifying effect (hexadecanol, octadecanol) on the membranes.¹⁶⁹ Unfortunately, the degree of fluidization produced by clinical concentrations of anesthetics is quite small.¹⁶⁸ While it is unclear how much fluidization would be required to affect ion channel function, anesthetics produce changes in membrane fluidity that can be mimicked by changes in temperature of less than 1°C. Clearly, a 1°C increase in temperature does not cause anesthesia, or even increase anesthetic potency. It is highly unlikely that changes in the fluidity of bulk membrane lipid are responsible for general anesthesia.

Lipid Phase Transitions

Another lipid perturbation that has been proposed to account for general anesthesia is a change in lipid phase-transition behavior. In its original version this theory proposed that anesthetics promote a transition of the lipids in neuronal membranes between a solid (gel) phase and a liquid-crystalline phase. Indeed, in pure lipid systems clinical concentrations of anesthetics do decrease the temperature at which such a transition occurs.¹⁷⁰ A second version of this theory, the lateral phase-separation theory, proposed that anesthetics prevent phase transitions between the liquid-crystalline and the gel phase.¹⁷¹ According to this theory, liquid-crystalline to gel phase transition is required for normal ion channel function; inhibition of this phase transition causes anesthesia. There is little evidence to support the phase-transition theories. Anesthetic-induced phase changes have not been observed in biologic membranes, lipid phase transitions are not known to be required for normal ion channel function, and the changes in phase-transition temperature observed in pure lipid systems are less than 1°C.

Protein Theories of Anesthesia

The Meyer-Overton rule could also be explained by the direct interaction of anesthetics with hydrophobic sites on proteins. Three types of hydrophobic sites on proteins might interact with anesthetics:

1. Hydrophobic amino acids comprise the core of water-soluble proteins. Anesthetics could bind in hydrophobic pockets that are fortuitously present in the protein core.

2. Hydrophobic amino acids also form the lining of binding sites for hydrophobic ligands. For example, there are hydrophobic pockets in which fatty acids tightly bind on proteins such as albumin and the low–molecular-weight fatty acid–binding proteins. Anesthetics could compete with endogenous ligands for binding to such sites on either water-soluble or membrane proteins.

3. Hydrophobic amino acids are major constituents of the α-helices, which form the membrane-spanning regions of membrane proteins; hydrophobic amino acid side chains form the protein surface that faces the membrane lipid. Anesthetic molecules could interact with the hydrophobic surface of these membrane proteins, disrupting normal lipid–protein interactions and possibly directly affecting protein conformation. This last possibility would involve the interaction of many anesthetic molecules with each membrane protein molecule and would probably be a nonselective interaction between anesthetic molecules and all membrane proteins.

Direct interactions of anesthetic molecules with proteins would not only satisfy the Meyer-Overton rule, but would also provide the simplest explanation for compounds that deviate from this rule. Any protein-binding site is likely to be defined by properties such as size and shape in addition to its solvent properties. Limitations in size and shape could reduce the binding affinity of compounds beyond the cutoff, thus explaining their lack of anesthetic effect. Enantioselectivity is also most easily explained by a direct binding of anesthetic molecules to defined sites on proteins; a protein-binding site of defined dimensions could readily distinguish between enantiomers on the basis of their different shape. Protein-binding sites for anesthetics could also explain the convulsant effects of some polyhalogenated alkanes. Different compounds binding (in slightly different ways) to the same binding pocket can produce different effects on protein conformation and hence on protein function. For example, there are three kinds of compounds that can bind at the benzodiazepine binding site on the GABA_A channel: agonists, which potentiate GABA effects and produce sedation and anxiolysis; inverse-agonists, which promote channel closure and produce convulsant effects; and antagonists, which produce no effect on their own but can competitively block the effects of agonists and inverse-agonists. By analogy, polyhalogenated alkanes could be inverse-agonists, binding at the same protein sites at which halogenated alkane anesthetics are agonists. The evidence for direct interactions between anesthetics and proteins is briefly reviewed in the following section.

Evidence for Anesthetic Binding to Proteins

One of the initial approaches to probing anesthetic interactions with proteins was to observe the effects of anesthetics on the function of a protein and to try to make inferences about binding from the functional behavior. It is entirely reasonable to assume that direct anesthetic-protein interactions are responsible for functional effects of anesthetics on purified water-soluble proteins because no lipid or membrane is present in the preparations studied. Firefly luciferase is a water-soluble, light-emitting protein, which is inhibited by a wide variety of anesthetic molecules. Numerous studies have extensively characterized anesthetic inhibition of firefly luciferase activity and have revealed the following:¹⁷²,¹⁷³

1. Anesthetics inhibit firefly luciferase activity at concentrations very similar to those required to produce clinical anesthesia.

2. The potency of anesthetics as inhibitors of firefly luciferase activity correlates strongly with their potency as anesthetics, in keeping with the Meyer-Overton rule.

3. Halothane inhibition of luciferase activity is competitive with respect to the substrate D-luciferin.

4. Inhibition of firefly luciferase activity shows a cutoff in anesthetic potency for both n-alkanes and n-alkanols.

Based on these studies it can be inferred that a wide variety of anesthetics can bind in the luciferin-binding pocket of firefly luciferase. The fact that anesthetic inhibition of luciferase activity is consistent with the Meyer-Overton rule, occurs at clinical anesthetic concentrations, and explains the cutoff effect suggests that the luciferin-binding pocket may have physical and chemical characteristics similar to those of a putative anesthetic binding site in the CNS.

More direct approaches to study anesthetic binding to proteins have included NMR spectroscopy and photoaffinity labeling. Based on early studies by Wishnia and Pinder, it was suspected that anesthetics could bind to several fatty acid-binding proteins, including y8-lactoglobulin and bovine serum albumin (BSA).¹⁷⁴,^{175 19}F-NMR spectroscopic studies confirmed¹⁷⁶ this, and demonstrated that isoflurane binds to approximately three saturable binding sites on BSA. Isoflurane binding is eliminated by co-incubation with oleic acid, suggesting that isoflurane binds to the fatty acid-binding sites on albumin. Other anesthetics, including halothane, methoxyflurane, sevoflurane, and octanol, compete with isoflurane for binding to BSA.¹⁷⁷ The studies with BSA provide direct evidence that a variety of anesthetics can compete for binding to the same site on a protein. Using this BSA model, it was subsequently shown that anesthetic binding sites could be identified and characterized using a photoaffinity labeling technique. The anesthetic halothane contains a carbon-bromine bond. This bond can be broken by ultraviolet light generating a free radical. That free radical allows the anesthetic to permanently (covalently) label the anesthetic binding site. Eckenhoff and colleagues used ¹⁴C-labeled halothane to photoaffinity-label anesthetic binding sites on BSA¹⁷⁸ and obtained results virtually identical to those obtained using NMR spectroscopy. Eckenhoff subsequently has identified the specific amino acids that are photoaffinity-labeled by [¹⁴C] halothane.¹⁷⁹ NMR and photoaffinity-labeling techniques have also been applied to several other proteins. For example, saturable binding of halothane to the luciferin-binding site on firefly luciferase has been directly confirmed using NMR and photoaffinity-labeling techniques.¹⁸⁰ Both NMR and photoaffinity-labeling techniques are also being applied to membrane proteins. At the current time these techniques can only be applied to purified proteins available in relatively large quantity. The muscle-type nicotinic acetylcholine receptor is one of the few membrane proteins that can be purified in large quantity. Eckenhoff has used photoaffinity labeling to show that halothane binds to this protein. The pattern of photoaffinity labeling is complex, suggesting multiple binding sites.¹⁸¹ Most recently, Miller and colleagues have developed a general anesthetic that is an analog of octanol and functions as a photoaffinity label. This compound, 3-diazyrinyloctanol, also binds to specific sites on the nicotinic acetylcholine receptor.¹⁸²

Although NMR and photoaffinity techniques can provide extensive information about anesthetic binding sites on proteins, they cannot reveal the details of the three-dimensional structure of these sites. X-ray diffraction crystallography can provide this kind of three-dimensional detail and has been used to study anesthetic interactions with a small number of proteins. To date, it has been difficult to crystallize membrane proteins; thus, these studies have been limited to water-soluble proteins. In 1965, Schoenborn and colleagues first used x-ray diffraction techniques to examine the interactions of several anesthetics with crystalline myoglobin.^183,184 These studies demonstrated that at a partial pressure of 2.5 atm (xenon MAC = 1 atm), a single molecule of xenon binds to a specific pocket in the hydrophobic core of the myoglobin molecule. The anesthetics cyclopropane and dichloromethane also bind in this pocket, but larger anesthetics do not. It should be noted that xenon occupies a small empty space in the hydrophobic core of myoglobin and that even dichloromethane is a tight fit in this space. These data provided a clear demonstration that anesthetic molecules can bind in the hydrophobic core of a water-soluble protein and that the size of the hydrophobic binding pocket can account for a cutoff in the size of anesthetic molecules that can bind in that pocket. However, myoglobin cannot bind most anesthetic molecules (because of their size) and is therefore not a good model for the actual anesthetic binding site(s) in the central nervous system.

X-ray diffraction has also been used to demonstrate that a single molecule of halothane binds in a hydrophobic pocket deep within the enzyme adenylate kinase.¹⁸⁵ Halothane binding was localized to the binding site for the adenine moiety of AMP (adenine monophosphate), a substrate for adenylate kinase. Consistent with this finding, halothane was found to inhibit adenylate kinase in a manner that is competitive with respect to AMP. Unfortunately, halothane binding to adenylate kinase only occurs at concentrations well beyond the clinically useful range. More recently, firefly luciferase has been crystallized in the presence and absence of the anesthetic bromoform. X-ray diffraction studies of these crystals showed that the anesthetic does bind in the luciferin-binding pocket, as had been inferred from functional studies. Interestingly, two molecules of bromoform bind in the luciferin pocket—one that is likely to compete directly with luciferin for binding and one that is not.¹⁸⁶ The binding data with firefly luciferase and adenylate kinase are of particular interest because they demonstrate that anesthetics can bind to endogenous ligand binding sites and that this binding strongly correlates with anesthetic inhibition of protein function. The same group has also crystallized human serum albumin in the presence of either propofol or halothane. The x-ray crystallographic data demonstrate binding of both anesthetics to preformed pockets that had been shown previously to bind fatty acids.¹⁸⁷ Given that both of these anesthetics bind to serum albumin at clinical concentrations, these data give the best insight yet into the structure of an anesthetic binding pocket.

A recent approach to study anesthetic interactions with proteins has been to employ site-directed mutagenesis of candidate anesthetic targets, coupled with molecular modeling to make predictions about the location and structure of anesthetic binding sites. For example, Harris, Trudell, and colleagues have used this approach to predict the location and structure of the alcohol binding site on GABA_A and glycine receptors.¹⁸⁸ A related approach has been to develop model proteins to

define the structural requirements for an anesthetic binding site. Using this approach, Johansson has shown that a four–α-helix bundle with a hydrophobic core can bind volatile anesthetics at concentrations (K_D) similar to those required to produce anesthesia.¹⁸⁹

Summary

Unequivocal evidence from studies using water-soluble proteins demonstrates that anesthetics can bind to hydrophobic pockets on proteins. Functional and binding studies with firefly luciferase demonstrate that anesthetics can bind to a protein site at clinically relevant concentrations in a manner that can account for the Meyer-Overton rule and deviations from it. Evidence that direct anesthetic–protein-binding interactions may be responsible for anesthetic effects on ion channels in the CNS remains indirect; stereoselectivity currently offers the strongest indirect argument.

Overall, current evidence strongly indicates protein rather than lipid as the molecular target for anesthetic action. While the long-standing controversy between lipid and protein theories of anesthesia may be behind us, numerous unanswered questions remain about the details of anesthetic–protein interactions including:

1. What is the stoichiometry of anesthetic binding to a protein? (i.e., Do many anesthetic molecules interact with a single protein molecule or only a few?)

2. Do anesthetics compete with endogenous ligands for binding to hydrophobic pockets on protein targets or do they bind to fortuitous cavities in the protein?

3. Do all anesthetics bind to the same pocket on a protein or are there multiple hydrophobic pockets for different anesthetics?

4. How many proteins have hydrophobic pockets in which anesthetics can bind at clinically used concentrations?

HOW ARE THE EFFECTS OF ANESTHETICS ON MOLECULAR TARGETS LINKED TO ANESTHESIA IN THE INTACT ORGANISM?

The previous sections have described how anesthetics affect the function of a number of ion channels and signaling proteins, probably via direct anesthetic-protein interactions. It is unclear which, if any, of these effects of anesthetics on protein function are necessary and/or sufficient to produce anesthesia in an intact organism. A number of approaches have been employed to try to link anesthetic effects observed at a molecular level to anesthesia in intact animals. These approaches and their pitfalls are briefly explored in the following section.

Pharmacological Approaches

An experimental paradigm frequently used to study anesthetic mechanisms is to administer a drug thought to act specifically at a putative anesthetic target (e.g., a receptor agonist or antagonist, an ion channel activator or antagonist), then determine whether the drug has either increased or decreased the animal's sensitivity to a given anesthetic. The underlying assumption is that if a change in anesthetic sensitivity is observed, then the anesthetic is likely to act via an action on the specific target of the administered drug. This is a largely flawed strategy that has nonetheless produced a huge literature. The drugs used to modulate anesthetic sensitivity usually have their own direct effects on central nervous system excitability and thus indirectly affect anesthetic requirements. For example, while a₂-adrenergic agonists decrease halothane MAC,¹⁹⁰ they are profound CNS depressants in their own right and produce anesthesia by mechanisms distinct from those used by volatile anesthetics. Thus, the “MAC-sparing” effects of a₂-agonists provide little insight into how halothane works. A more useful pharmacological strategy would be to identify drugs that have no effect on CNS excitability but prevent the effects of given anesthetics. Currently, however, there are no such anesthetic antagonists. Development of specific antagonists for anesthetic agents would provide a major tool for linking anesthetic effects at the molecular level to anesthesia in the intact organism, and might also be of significant clinical utility.

An alternative pharmacological approach is to develop “litmus tests” for the relevance of anesthetic effects observed in vitro. One such test takes advantage of compounds that are nonanesthetic despite the predictions of the Meyer-Overton rule. It is argued that “a site affected by these nonanesthetic compounds is unlikely to be relevant to the production of anesthesia.”¹⁴⁸ A similar argument uses stereoselectivity as the discriminator and argues that a site that does not show the same stereoselectivity as that observed for whole animal anesthesia is unlikely to be relevant to the production of anesthesia.¹⁹¹ Although these tests may be useful, they are very dependent on the assumption that anesthesia is produced via drug action at a single site. For example, a nonanesthetic might depress CNS excitability via its actions on an important anesthetic target site while simultaneously producing counterbalancing excitatory effects at a second site. In this case the “litmus test” would incorrectly eliminate the anesthetic site as irrelevant to whole animal anesthesia. This example is quite plausible given the convulsant effects of many of the nonanesthetic polyhalogenated hydrocarbons. Another sort of litmus test is to selectively antagonize the putative anesthetic target so that this target is no longer functional. If anesthetic effects are mediated through this target, inactivation of the target by the antagonist should result in anesthetic resistance. Using this logic, the MAC-sparing effects of GABA_A and glycine receptor antagonists were used to argue that both GABA_A and glycine receptors mediate some but not all of the immobilizing effects of volatile anesthetics in rodents.^192,193 This same group used the lack of effect of neuronal nicotinic antagonists on isoflurane MAC to conclude that these receptors had no role in volatile anesthetic immobilization.¹²⁷ As with many pharmacological results, the issues of specificity and efficacy of the antagonists prevent these experiments from being definitive. Nevertheless, these results are consistent with the findings that volatile anesthetics affect the function of a large number of important neuronal proteins and no one target is likely to mediate all of the effects of these drugs.

Genetic Approaches

An alternative approach to study the relationship between anesthetic effects observed in vitro and whole animal anesthesia is to alter the structure of putative anesthetic targets and determine how this affects whole animal anesthetic sensitivity. Genetic techniques provide the most reliable and versatile methods for changing the structure of putative anesthetic targets. Toward this end, a variety of approaches have been taken that can be methodologically categorized as selective breeding, forward genetics, and reverse genetics. Selective breeding makes use of existing genetic variance among strains that are presumably because of differences in multiple genes and attempts to breed and select for enhanced differences in the trait of interest—in this case general anesthetic sensitivity. Koblin and colleagues have successfully used this strategy to breed two strains of mice (HI and LO) that differ in their sensitivity to N₂O by almost 1.0 atm.¹⁹⁴ A similar strategy has been used to breed mice that have differential sensitivity to the hypnotic effects of the benzodiazepine, diazepam. The two strains of mice (DR and DS) show some modest, but consistent, differences in their sensitivity to volatile anesthetics.¹⁹⁵ Both sets of strains have differences in sensitivities to drugs other than general anesthetics;¹⁹⁶ thus, it seems likely that the genetic differences in these strains may be more general differences in brain function/excitability rather than specific differences in an anesthetic target. Nevertheless, the HI/LO and DS/DR strains demonstrated that in principle genes controlling anesthetic sensitivity, perhaps encoding anesthetic targets, could be discovered. These strains have not led as yet to the identification of the responsible genetic loci. Even under the best of circumstances, mapping genes to the point of their identification in mice is exceedingly difficult, time-consuming, and expensive. In the particular cases of mapping the anesthetic sensitivity loci in these strains, the task is made even more difficult because the phenotype being mapped requires special testing and there is overlap in anesthetic sensitivity between the strains. Further, at least for the HI/LO strains, multiple genetic loci are contributing to the differences in anesthetic sensitivity.¹⁹⁶ Multiple loci are much more complex to identify because typically the contribution of each to the phenotype is small and therefore easily lost in the environmental noise. Forward genetics refers to the classical approach of starting from a phenotype of interest, for example, altered anesthetic sensitivity, and moving “forward” ultimately to identify the gene of interest. Strictly speaking, selective breeding is one form of forward genetics although it rarely proceeds all the way to identification of the genes responsible for the phenotype. More commonly, forward genetics involves inducing random mutations throughout the genome of a pool of animals, then identifying the rare individual that carries a mutation producing the phenotype of interest. This approach requires screening through large numbers of animals and can be effectively accomplished only in lower organisms with a large number of offspring and short generation times. This typically means invertebrate models such as the fruit fly or nematode.

The first true forward genetic screen for mutants with altered general anesthetic sensitivity was performed in the nematode C. elegans by Phil Morgan and Margaret Sedensky.^{197 198} They screened for altered sensitivity to supraclinical concentrations of halothane. High halothane concentrations were used because they are required to immobilize C. elegans. The first mutant isolated had a three-fold reduction in its EC₅₀ for halothane. Interestingly, this mutant was hypersensitive to chloroform, methoxyflurane, and thiomethoxyflurane but not to less lipophilic anesthetics such as isoflurane and enflurane.¹⁹⁸ This selective hypersensitivity argues that a generalized nervous system dysfunction is unlikely to account for the halothane hypersensitivity. The mutation was genetically mapped and found to be a loss-of-function allele of the unc-79 gene, which encodes a neuronal protein that is most similar in amino acid sequence to a large human protein encoded by a gene on chromosome 14.¹⁹⁹ The cellular function of either the C. elegans or human protein is unknown. To attempt to understand the function of unc-79, a search for mutations that return the halothane hypersensitivity of the unc-79 mutants toward normal levels was undertaken. Mutations in genes encoding stomatin-like proteins, an integral membrane protein first identified in erythrocytes, were found to suppress unc-79.²⁰⁰ Genetic evidence suggested that the C. elegans stomatins might control halothane sensitivity by regulating the function of a mechanically gated sodium channel.²⁰¹ Additional mutant screens identified a gene, called gas-1, which encodes a highly conserved mitochondrial protein functioning in the electron transport chain.²⁰² gas-1 mutants were hypersensitive to all halogenated volatile anesthetics tested. The mechanistic relationship between gas-1 and unc-79 and its suppressors genes is unclear.

Clinical concentrations of volatile anesthetics do not immobilize C. elegans, but they do produce behavioral effects including loss of coordinated movement.²⁰³ Crowder and colleagues have screened for mutants that are resistant to anesthetic-induced uncoordination and found that mutations in a set of genes encoding proteins regulating neurotransmitter release control anesthetic sensitivity. The gene with the largest effect encoded syntaxin 1A, a neuronal protein highly conserved from C. elegans to humans and essential for fusion of neurotransmitter vesicles with the presynaptic membrane.⁵⁴ Importantly some syntaxin mutations produced hypersensitivity to volatile anesthetics while others conferred resistance. These allelic differences in anesthetic sensitivity could not be accounted for by effects on the process of transmitter release itself;⁵⁴⁵⁵ rather, the genetic data argued that syntaxin interacts with a protein critical for volatile anesthetic action, perhaps an anesthetic target. This putative target has not yet been identified.

In Drosophila, clinical concentrations of volatile anesthetics disrupt negative geotaxis behavior and response to a noxious light or heat stimulus.²⁰⁴,²⁰⁵,²⁰⁶ Using one or more of these anesthetics effects, Nash and colleagues performed a forward genetic screen for halothane resistance. Several har (halothane resistance) mutants were isolated. One set of mutants, har38 and har85, was found to have mutations in a gene encoding a putative cation channel with predicted structural similarities to both sodium and calcium channels.²⁰⁷ Interestingly, halothane was shown to reduce glutamatergic transmission at the Drosophila larval neuromuscular junction, most likely by inhibiting glutamate release, and the har38 and har85 mutants were resistant to this presumed presynaptic halothane action.²⁰⁸ As the identification of the syntaxin mutants suggested in C. elegans, this result suggests that inhibition of excitatory neurotransmitter release may be a consequential action of volatile anesthetics in disrupting behavior in Drosophila.

At anesthetic concentrations 1.5- to 2-fold higher than MAC, volatile anesthetics ablate response of Drosophila to touch.²⁰⁹ Using this anesthetic end point, Gamo and colleagues have extensively screened for Drosophila mutants with altered sensitivity to diethyl ether. Mutated genes in two of the strains have been identified. A partial-loss-of-function mutation in the a subunit of the major neuronal sodium channel mediating action potentials (para) was one of the mutants. This para Na ⁺ channel mutant had about a 50% reduction in its ether EC₅₀.²¹⁰,²¹¹ A mutation in the Drosophila calreticulin gene was also found to produce similar hypersensitivity to ether.²¹² Interestingly, this calreticulin mutant was mildly resistant to isoflurane and normally sensitive to halothane.

Calreticulin is a highly conserved protein localized to the endoplasmic reticulum of all cell types and is involved in Ca²⁺ buffering and protein folding in the ER.²¹³ Because of this broad role in cellular function, calreticulin's role in anesthetic sensitivity could be indirect.

As with all model organisms, a critical question to ask is how do the anesthetic mechanisms implicated in nematode and fruit fly relate to mechanisms of anesthesia in humans? Even if a similar gene exists in humans, the evolutionary divergence of the molecules and the very different nervous systems makes the relevance question impossible to answer without additional experiments. Thus, a more practical question is which of the implicated invertebrate genes deserves a potentially more arduous and expensive examination in a vertebrate species? A few criteria seem reasonable. First, is the gene involved in a process known to be affected by general anesthetics in vertebrates? Certainly, the genes in C. elegans and Drosophila encoding proteins regulating neurotransmitter release and the sodium channel fit this criterion. While mitochondrial electron transport has generally not been implicated in vertebrate anesthetic action, a case report of four children who are hypersensitive to sevoflurane by processed EEG criteria and found to carry defects in the same mitochondrial protein complex implicated in C. elegans is an intriguing observation that would likely not have been made without the work in C. elegans.²¹⁴ Second, is the gene conserved in vertebrates and does it function in the nervous system? In this regard, both the mitochondrial protein and syntaxin 1A are very highly conserved and both function in the nervous system with syntaxin 1A expressed exclusively in neurons; however, for each of these proteins one must explain the enigma of neuron-subtype-specific effects of anesthetics by a protein that functions in all neurons. A third criterion is anesthetic concentration. Do the genes regulate sensitivity to clinical concentrations of anesthetics? However, in this case, some latitude should be given for the possibility that the binding sites on the anesthetic targets are partially diverged and therefore the affinity of the target could be reduced. Certainly, experiments with GABA_A receptors and model anesthetic binding proteins have shown that single amino acid changes can drastically alter anesthetic potency or affinity.¹¹⁶, 215,216,217,218 Thus, anesthetic concentration criteria should not be used to exclude mechanisms as is reasonably done in more closely related species such as rodents; rather, the “correct concentration” neither rules in or out the mechanism in question but simply makes it more plausible. Finally, one should keep in mind that even if a particular anesthetic mechanism identified in invertebrates is operant in humans, it may not be the only mechanism of anesthetic action in humans and indeed it may not even be involved in anesthesia at all but rather in anesthetic side effects in other tissues such as myocardium or vascular smooth muscle. Thus, invertebrate genetics should be viewed as a means to pose novel hypotheses, some of which may be compelling enough to test in vertebrates.

Reverse genetics refers to altering the sequence of a known gene and then observing the effects of this mutation on the process of interest. In other words, reverse genetics moves from gene to phenotype as opposed to classical forward genetics that starts with a phenotype and then proceeds to identify the responsible gene(s). Reverse genetics is used typically to test a well-established hypothesis, although occasionally surprising phenotypes produce novel hypotheses. While reverse genetics is employed in both invertebrate and vertebrate models, in terms of anesthetic sensitivity, reverse genetics has been most instructive in mice.

The GABA_A receptor has been extensively studied using reverse genetic techniques.²¹⁹ The genes encoding for various subunits of the GABA_A receptor have been mutated so that they are either nonfunctional gene knockouts) or so that they have altered amino acids that might produce altered function (gene knockins). Knockouts of two α subunits of the GABA_A receptor have been tested for their anesthetic sensitivity. Lack of the α1 subunit was not found to alter sensitivity of the animal to the hypnotic effects of pentobarbital.²²⁰ Similarly, α6 subunit knockout mice were normally sensitive to halothane and enflurane.²²¹ Knockin mouse strains have been generated for several of the a-subunits, primarily for examining benzodiazepine action. The loss of various aspects of benzodiazepine action in these strains demonstrated that the α1 subunit mediates the sedative and amnestic actions, and is partially required for its anticonvulsant properties. Similarly, the α2 subunit was found to be essential for anxiolysis by diazepam, and α3 and α5 knockin strains were partially resistant to its myorelaxant effects. However, none of these a-subunit knockin strains have been reported to be abnormally sensitive to any complete general anesthetics. In contrast, knockout of the β3 subunit produced mice with a markedly decreased sensitivity to the hypnotic action of both midazolam and etomidate and a mildly decreased sensitivity to halothane and enflurane in tail clamp response assays.²²² The interpretation of these data was complicated by a variety of behavioral and neurological abnormalities in these mice that suggested the possibility of an indirect effect of the mutation on anesthetic sensitivity.

In vitro electrophysiological experiments had shown that a specific β3 subunit point mutation, β3 (N265M), blocked the action of etomidate and propofol on the GABA_A receptor without greatly altering receptor function in the absence of drug;¹¹⁷²²³ this result suggested a means to circumvent the problems produced by knocking out β3. Thus, a mouse β3(N265M) knockin strain was generated and found to be insensitive to the immobilizing effects of etomidate and propofol.²²⁴ However, the β3(N265M) mice were not completely resistant to the loss-of-righting reflex by etomidate and propofol, indicating that other targets mediated this behavioral effect. Volatile anesthetic sensitivity was modestly reduced in the β3(N265M) mice suggesting that the β3 subunit may play some role in their action. A similar approach was taken to show that the β2 subunit is critical for the sedating but not anesthetic action of etomidate.²²⁵,²²⁶ Finally, strains carrying a knockout mutation of the δ subunit of the GABA_A receptor were found to have a shorter duration of neurosteroid-induced loss-of-righting reflex, whereas their sensitivity was normal to other intravenous and volatile anesthetics.²²⁷ Thus, the δ subunit may play a relatively specific role in neurosteroid action.

Summary

Overall, genetic studies provide a powerful tool for determining which genes and gene products are important in producing anesthesia in an intact organism. Forward genetics has the potential to identify anesthetic mechanisms/targets that may not have been implicated by vertebrate biochemical and electrophysiological studies that are biased toward abundant ion channels. However, particularly for invertebrate genetics, the genetically identified mechanisms may not be operant in humans or may be operant in a different physiological context. Reverse genetics has strengths and weaknesses complementary to those of forward genetics. Reverse genetics rarely generates novel hypotheses or fundamental breakthroughs, but it can confirm definitively the in vivo role of a gene product. Indeed, the demonstration that the action of the general anesthetics etomidate and propofol can be blocked by a single missense mutation in a subunit of the GABA_A receptor is at the same time not surprising and yet one of the most important results thus far in anesthetic mechanism research.

CONCLUSION

In this chapter evidence has been reviewed concerning the anatomic, physiologic, and molecular loci of anesthetic action. It is clear that all anesthetic actions cannot be localized to a specific anatomic site in the central nervous system; indeed, some evidence suggests that different components of the anesthetic state may be mediated by actions at disparate anatomic sites. The actions of anesthetics also cannot be localized to a specific physiologic process. While there is consensus that anesthetics ultimately affect synaptic function as opposed to intrinsic neuronal excitability, the effects of anesthetics are dependent on the agent and synapse studied and can affect presynaptic and/or postsynaptic function. At a molecular level, anesthetics show some selectivity, but still affect the function of multiple ion channels and signaling proteins. Although it is likely that these effects are mediated via direct protein-anesthetic interactions, it appears that there are numerous proteins that can directly interact with anesthetics. All of these data suggest that the unitary theory of anesthesia is incorrect and that there are at least several molecular mechanisms of anesthesia.

In keeping with the idea that anesthesia can be produced in a variety of ways, Pancrazio and Lynch have suggested that different anesthetic targets may mediate different components of the anesthetic state.²²⁸ As illustrated in Fig. 8, they suggest that the analgesic effects of opiates, α₂-agonists, and volatile anesthetics are mediated via inhibition of calcium currents and/or activation of potassium currents. Sedation and amnesia, they propose, are mediated by potentiation or activation of GABA_A receptors. In this model, anesthetic states can also be produced by totally independent mechanisms such as the inhibition of glutamate receptors by ketamine. Although there may be many more important anesthetic targets than those suggested by Pancrazio and Lynch, their proposal illustrates the idea that different molecular targets may mediate the various components of the anesthetic state, and that volatile anesthetics are complete anesthetics because they can interact with several of these molecular targets.

FIGURE 8. A multisite model for anesthesia. The model proposes that presynaptic (Ca2+- analgesic effects, whereas postsynaptic GABAA receptor activation is responsible for sedation and amnesia. As indicated by the overlapping circles, the behavioural effect of Ca2+A receptor activation are not mutually exclusive. The model suggest that some anesthetic agents predominantly affect Ca2+ and K+ channel, other anesthethic agents predominantrly affect GABAA­ receptors, and volatile anesthetics affect both. As illustrated at the bottom of the model, inhibition of glutamate receptor function is an alternative pathway by which ketamine and perhaps the volatile anesthetics produce anesthesia. (Reproduced with permission from Pancrazio JJ, Lynch C. Snails, spiders, and stereospecificity—Is there a role for calcium channels in anesthetic mechanisms? Anesthesiology. 1994;81:3.)

Although the precise molecular interactions responsible for producing anesthesia have not been fully elucidated, it has become clear that anesthetics do act via selective effects on specific molecular targets. The technologic revolutions in molecular biology, genetics, and cell physiology make it likely that the next decade will provide the answers to the century-old pharmacological puzzle of the molecular mechanism of anesthesia.