Chapter 37. Mechanisms of General Anesthetic Action

1. The mechanisms by which the inhaled general anesthetics work are not fully understood. No single molecular target has been proven to transduce anesthesia.

2. Correlation of the physicochemical character of anesthetics with their potency suggests that target sites are dominantly hydrophobic, with a small degree of polarity and chirality. Internal or interfacial protein cavities best fit this description.

3. Inhaled anesthetic binding site character is not highly specific, predicting more than a few anesthetic binding targets. The interaction at some of these targets may not contribute significantly to anesthetic action but may contribute to side effects.

4. Use of the lipid membrane as a direct target for inhaled anesthetics has been dismissed prematurely. Some components of anesthetic action may occur via this interaction.

5. Many potential protein targets in the synapse have been identified, suggesting that inhaled anesthetic action results from disruption of the specific process of synaptic transmission rather than from a receptor-like interaction with a single molecular target.

6. Anesthetic effects on a process, such as synaptic transmission, may have a different system-level effect depending on placement in the neural circuitry. The circuits that regulate sleep and arousal are well positioned to mediate the hypnotic properties of anesthetics.

General anesthetics were formally introduced into medical practice more than 160 years ago and have been hailed as one of the most significant medical advances of all time. Yet there is still considerable mystery about how the drugs work and how to characterize the state that they produce. This should not be too surprising because a description of the transition from consciousness to unconsciousness necessarily requires an understanding of the former, and the neurobiology of consciousness is still in its infancy. Nevertheless, considerable progress has been made toward the characterization of anesthesia and the potential mechanisms of the drugs that produce it. In this chapter, we summarize the current body of evidence for the mechanism(s) of general anesthesia, with emphasis on the inhaled anesthetics because they are used most commonly. Excellent and comprehensive reviews summarize current knowledge of the molecular, cellular, and in vivo pharmacology,1,2 so we have selected only a few of the putative molecular targets to illustrate the principles and to support the notion that alteration of a neurophysiologic process, rather than the activity of an individual protein, is responsible for the state of general anesthesia.

Mechanistic searches are greatly aided by clear physiologic or behavioral endpoints. Those associated with anesthesia, however, often are ambiguous and arbitrary. Most people associate anesthesia with unconsciousness or "sleep," but the transition to this state often is not apparent to the observer, so the lack of physical movement in response to a noxious stimulus is the most common endpoint associated with the term anesthesia. This is essentially the same endpoint used in Boston more than 160 years ago. But it is immediately apparent that many pathways can lead to such an endpoint, most notably paralysis, which itself is at the end of many pathways. Nevertheless, the mobility endpoint is well entrenched and has been useful as a practical comparison of the potency of different drugs. After all, surgery is greatly facilitated by a motionless patient. Thus was born the concept of minimum alveolar concentration (MAC; as described in this chapter), a useful tool for practitioners. Investigators, on the other hand, were hampered by such a concept and only now are starting to break the state of anesthesia into its many components in an attempt to relate them to the underlying targets and pathways. For example, most acknowledge the presence of hypnosis, amnesia, excitement, weakness, and analgesia in the state of anesthesia. These are distinct from the many arbitrarily assigned "side" effects of the drugs, including alterations in vascular, cardiac, respiratory, metabolic, and renal function. Also produced is a series of "toxic" effects, which might be unique or simply an enhancement of the primary or the side effects. This disassembly of effects is likely to facilitate the linkage with molecular targets and with their reassembly to the final in vivo state of anesthesia. At the very least, it facilitates the formation and testing of critical hypotheses. We return to these ideas toward the end of this chapter.

Mechanistic searches are also aided by clear structural motifs among the drugs that produce the endpoint. This permits a search for the complementary motif on the target macromolecules. Here too, general anesthetics fall short of the mark. A structurally diverse range of molecules is capable of producing a state of anesthesia (Fig. 37-1). Nonetheless, certain physicochemical features appear to be important. For example, hydrophobicity correlates exceedingly well with anesthetic potency, with only a few exceptions (Fig. 37-2). Although impressive, this only suggests that the domain(s) producing the effects is hydrophobic, limiting the choices to most proteins and all lipid membranes. Other more subtle features of the drugs seem to be important. Most inhaled anesthetics have an asymmetric distribution of hydrogen atoms, giving the molecule enough of a dipole moment to interact favorably with polarity within the hydrophobic environments. The few nonimmobilizer drugs (Fig. 37-3), introduced to allow for selection of relevant targets,3,4 lack this feature and therefore are extremely insoluble in water.4 Although the mechanism of their failure to cause anesthesia is not clear, it likely is related to their low dipole and poor solubility in water and therefore lower occupancy in targets. Also, there is a clear relationship between potency and the presence of halogens (and their size) on the anesthetic molecule. Given that the halogens are uncharged atoms, then polarizability must be important, another indication that the sites responsible for anesthesia have a degree of polarity. Despite a narrowing focus on protein as the molecular target (see below), lipids still cannot be excluded because the head group region has both polar and hydrophobic character.

Related to hydrophobicity is the cutoff effect (Fig. 37-4). In a homologous series of compounds, for instance, n-alcohols, anesthetic potency increases progressively as the hydrocarbon chain lengthens until a certain length is reached; then anesthetic potency is abruptly lost. This typically occurs at 10 to 14 carbons, depending on assay and endpoint, and has been used to implicate specific molecular targets.5 Interpretation of this phenomenon has varied, but is conventionally viewed as indicating the geometric capacity of a binding site—a molecular ruler. Of course, this interpretation requires that the alcohol actually bind to the target in question, a requirement that is almost never tested. In the 1 case in which it was, a cutoff effect was observed with no apparent loss in binding affinity,6 suggesting that cutoff is attributable to mechanisms other than steric hindrance, perhaps (similar to the nonimmobilizers) to the extremely low solubility of the long compounds in water. The final blow to the cutoff effect being useful as a molecular ruler of important protein binding sites was the demonstration that polyhydric alcohols (essentially n-alcohol polymers) retained anesthetic activity at molecular sizes in large excess of the n-alkanol cutoff.7

The final drug feature that has been used to aid the search for targets is enantioselectivity. Many of the inhaled anesthetics have chiral carbons, indicating that mirror image molecules are possible (Fig. 37-5). Such pairs are identical from a physicochemical standpoint but have a different arrangement of atoms in space, which would be expected to interact with a unique spatial distribution of atoms in a protein binding site, for example, more specifically than in a more fluid-like assembly of lipid molecules. Small differences in the immobilizing potency of the isoflurane enantiomers were found8 and in at least 1 species for the halothane enantiomers.9 Although this has the potential for selecting relevant targets, the small degree of enantioselectivity (˜20%), combined with the fact that many proteins display enantioselectivity for inhaled anesthetics, renders the potential small. It is even possible that the few chiral molecules in phospholipid bilayers (ie, cholesterol) or that very general and conserved chiral features of proteins (eg, L-α-amino acids and uniquely handed helices) contribute to this small degree of selectivity.

Thus, in general, drug features have given us only general clues to the identity of important anesthetic targets and the underlying mechanisms of their dysfunction.

Lipid Bilayers

The dramatic correlation of hydrophobicity with anesthetic potency predated an understanding of protein interiors, so initial focus fell on the obvious oily candidate: the lipid bilayer (Fig. 37-6). The idea was that solubilization of the hydrophobic anesthetic into the thin lipid membrane altered a property sensed by the cell or by proteins wholly or partially embedded in the lipid. The crucial property was (and still is) unclear, so surrogate properties that could be easily measured, such as the gel–liquid phase transition, were studied extensively. Changes induced by the anesthetics in these properties could be demonstrated, but several features gradually emerged that dampened enthusiasm in the lipid membrane as an important transducer of anesthetic effects. First, measurable changes in phase transition at clinical concentrations of anesthetic were small, mimicked by only 1° to 2°C. Second, other compounds that dissolved well and produced similar effects in these bilayers (eg, nonimmobilizers) did not produce anesthesia. Finally, the emergence of protein-centered theories shifted attention away from the lipid bilayer.

A shift in attention, however, does not suffice to eliminate the potential for an important contribution. New theories, coupled with improved understanding of the lipid bilayer and its interaction with anesthetics and membrane protein, call for a return to the lipid bilayer at some point to test whether anesthetic modulation contributes to in vivo anesthetic effects. For example, it is clear that phase transitions are not a relevant feature of biologic membranes, so conclusions based on their measurement must be viewed cautiously. It now is clear that anesthetics distribute in the lipid bilayer heterogeneously, less in the center (acyl trough) and more toward the outer edges. This asymmetric distribution can influence lipid properties not easily measured, such as lateral pressure (Fig. 37-7) but posited to have a potent effect on membrane protein conformational transitions.10 Interestingly, even though nonimmobilizers partition well into the lipid bilayer, the distribution within the bilayer is very different than that of an anesthetic,11 yet again opening the door for lipid theories. Furthermore, lipid microdomains exist (eg, lipid rafts) that appear to organize specific proteins into functional groups may have a higher affinity for inhaled anesthetics (because of unique lipids or unique proteins), increasing the opportunity for anesthetic effects on signaling pathways.12 Finally, in most assays of membrane protein activity (eg, ion channels), protein is functionally inseparable from the lipid. Cholesterol, for example, may contribute structurally essential roles in some membrane proteins by binding deep within the protein matrix.13 Therefore, mutagenesis of membrane protein by itself cannot be directly used to implicate a protein target because the mutation also might influence the interaction with lipid. Thus, it is premature to rule out the biologic lipid membrane as an important direct target for the inhaled anesthetics.

Proteins

Proteinaceous components of the cell as targets for anesthetics were first proposed by Claude Bernard in 1875 but did not gain favor until the demonstration that the activity of lipid-free preparations of protein were reversibly influenced by anesthetics and in a way that reproduced the correlation with hydrophobicity.14 Because this paralleled the understanding that the interior of proteins is as hydrophobic as that of lipid membranes, this association should not have been surprising. When combined with enantioselectivity and an ability to measure and alter protein function, especially in ion channels, protein-centered theories rapidly became favored. Many protein targets have been proposed and studied, and we review some of them briefly. However, at this time it is safe to conclude that not a single protein target or even class or family of protein targets has been proven to play the central role in inhaled anesthetic action. This is not to say, however, that favored candidates have not emerged, especially for the injectable induction agents. We first review some essential basics of protein–ligand interactions as they relate to anesthetics and then present the evidence for specific targets.

Binding

The submolecular mechanism by which the anesthetic causes a change in protein function or activity must first involve a binding event—formation of drug–target complex. This often is the initial criterion for establishing the relevance of any proposed drug target but has not been used for inhaled anesthetics because the affinity for protein targets is low, and they appear to be fairly promiscuous, binding to many targets. Both nuclear magnetic resonance spectroscopy and photolabeling have suggested a large number of protein binding targets (ie, 27) and no particular regional preference in the brain (Fig. 37-8).15,16 This renders untenable the conventional radioligand binding approach for discovering targets. Whether anesthetics also significantly alter the function of a large number of proteins is not clear, but even if only a minority of the binding target pool, the number of contributing targets is still large.

Nature of the Binding Site

Until recently, the nature of anesthetic binding sites has been inferred from correlations with anesthetic structural or physicochemical features. For example, the Overton-Meyer relationship suggested that sites are in the hydrophobic interior of proteins, and the relationship of halogen size and asymmetric hydrogens with potency suggested that polar amino acids must contribute to these sites. High-resolution structures of a clinically used inhaled anesthetic complexed to a protein target have been revealed for at least 3 protein models: serum albumin,17 apoferritin (Fig. 37-9),18 and integrin lymphocyte function-associated antigen 1 (LFA-1).19 Although unlikely to transduce desirable effects of anesthesia, the affinity of these proteins (see later discussion) suggests substantial occupancy at clinical concentrations of anesthetic (˜0.2 mmol/L), and thus the binding site architecture is likely to bear resemblance to those in targets underlying effects relevant to anesthesia. These structures show occupancy of preexisting internal packing defects, also known as cavities. The apoferritin example goes further to suggest that a packing density of just over half full represents an optimal balance between enthalpic and entropic forces. Finally, all of these examples suggest that polar (but uncharged) amino acids contribute to the strength of binding. In the apoferritin cavity, 2 serine–tyrosine–leucine triads interact with either halothane or isoflurane atoms to produce the highest affinity binding yet described for these compounds (Fig. 37-9). The stringency of these features is not yet clear, but preliminary estimates based on current entries in the Protein Data Bank (http://www.rcsb.org/pdb) suggest that less than 1% of the proteome (˜3000 proteins) have comparable binding sites for volatile anesthetics.

Affinity and Stoichiometry

Affinity of a drug for a site, indicated by the association constant (Ka) or, more commonly, the dissociation constant (Kd = 1/Ka), is experimentally defined as the drug concentration at which half the sites are occupied. It is more properly called "apparent" affinity because other aspects of the mixture can modify the true affinity of the ligand–site complex. Relevant to this is whether the binding event "causes" a conformational change (see later discussion) in the protein to which it binds. If so, the apparent affinity will be lowered by the amount of free energy used to change the protein conformation (Fig. 37-10). Thus, a set of binding site features that produces a 10 μmol/L absoluteKd with an anesthetic perhaps will have an apparentKd of 1 mmol/L if approximately 2 kcal/mol of free energy is required to change the protein conformation.

The slope of the relationship between drug concentration and occupancy, the Hill slope, is an indication of both the number of sites on a target and whether they interact. For example, a single site should always have a Hill slope of 1, reflecting a relationship in which the site goes from 0% to 100% occupancy over an approximately 100-fold change in drug concentration (Fig. 37-10). Multiple sites on a single target also can demonstrate a Hill slope of 1, or they can "cooperate," increasing the slope. In other words, if occupancy of the first site enhances occupancy of the second (as in oxygen binding by hemoglobin), the slope will rise, although rarely over 2 or 3. Although difficult to measure as indicated, anesthetic Kd values in model proteins typically are approximately 1 mmol/L, and Hill slopes are generally 1 or slightly greater even though more than 1 binding site has been found in most proteins studied. Serum albumin has at least 6 sites,17 apoferritin 12,18 and the intact nicotinic acetylcholine receptor (nAChR) 15 to 20.20-23 As might be deduced from the preceding paragraph, occupancy of multiple sites is a potentially powerful method of donating free energy to the protein to alter its conformation. Occupancy of only 1 additional binding site in a protein contributes as much free energy as an approximately 10-fold increase in affinity at a single site.

An issue frequently causing confusion has been what affinity an important anesthetic target should display for anesthetics. Prevailing logic has suggested that for a target to be relevant to a particular endpoint (eg, immobility), the dissociation constant for the anesthetic–target complex should approximate that achieved at MAC. However, this logic holds only if one assumes that this target is alone capable of producing the endpoint, an assumption that rarely is valid. Furthermore, even if the target is wholly responsible for the endpoint, it is difficult to know the expected Kd because drug efficacy at this target is not known. It is rare for the relationship between receptor occupancy and median effective concentration (EC50) to be linear; thus, predicting more than that some degree of occupancy should be anticipated at clinical concentrations is difficult. As our discussion on Hill slopes indicates, this limits the expected Kd to an approximately 100-fold range.

Coupling Mechanism

As discussed earlier, the anesthetic can only alter the protein's activity through a contribution of binding energy—through either the affinity or the stoichiometry of the complex. This binding energy alters protein activity through at least 3 potential and partially overlapping mechanisms. The first mechanism, competition, occurs when the drug–protein complex is strong enough to compete with the binding of an endogenous ligand, therefore inhibiting the associated effect (Fig. 37-11). This was the mechanism initially proposed for anesthetic inhibition of firefly luciferase.14 There are 2 forms of competition: isosteric and allosteric. In isosteric competition, the drug occupies the same binding site as the endogenous ligand, physically preventing binding. An anesthetic example is halothane occupancy of the retinal cavity in the G-protein coupled receptor (GPCR) rhodopsin.24 In allosteric competition, the drug binds elsewhere in the protein, altering the structure sufficiently to disfavor ligand binding in its otherwise unoccupied site. A clear example is integrin LFA-1, in which isoflurane binds in a cavity that stabilizes a conformation with low affinity for the intercellular adhesion molecule (ICAM) receptor protein.19 In a closely related form of allosterism, cooperativity, allosteric binding of anesthetic favors a conformation that binds the endogenous ligand more tightly. Thus, in firefly luciferase, the binding of adenosine triphosphate (ATP) and anesthetic is cooperative.25 Cooperative effects of agonist and anesthetic in some receptors may also be attributable to binding cooperativity.22,26

The second general category of drug–effect coupling is ensemble modulation (Fig. 37-12). Functional proteins exist in an ensemble of conformations, each of which is associated with some aspect of activity. Resting, active, desensitized, and so on all describe states of activity with a specific underlying protein conformation. Some conformers may possess binding sites for anesthetics that are more attractive than others and therefore are populated to a greater extent when the anesthetic is present. The anesthetic changes protein activity by selecting the most favorable conformer for binding and increasing its stability and therefore prevalence. One can immediately see the overlap with allosterism. Similar to allosterism, ensemble modulation can explain either an enhancement or an inhibition in protein activity, depending on which conformer has the most attractive anesthetic binding sites.

The final general mechanism for coupling binding to a change in protein function is oligomerization modulation (Fig. 37-13). Much protein activity and signaling is controlled by interactions with other proteins. In general, these interactions are highly specific but relatively weak so that they can be readily reversed. The protein–protein interface may include features attractive to an anesthetic, such as cavities. If these features are optimal when the proteins are linked, then the anesthetic will enhance the interaction or oligomerization; if they are optimal when separated, anesthetics will disfavor oligomerization. Examples of the former appear to be the sarcoplasmic reticular calcium adenosine triphosphatase (ATPase)27 and of the latter the PSD-95 PDZ domain proteins (see later discussion).28 The synaptic vesicle machinery is highly dependent on these oligomerization events, and components that might be modulated by anesthetics in this way have been identified.29 It should be apparent that this mechanism has important implications as to the number of potential effects these drugs might have. Whereas the set of all proteins expressed—the proteome—is very large (˜300 000 proteins), the number of protein–protein interactions, the so-called interactome, is much larger. Thus, anesthetic effects at the interactome level produced by effects at the interface could dwarf those at the individual protein level.

An important caveat in this discussion is that all binding interactions do not necessarily result in an important change in protein activity.30 It is possible that an anesthetic binding site is precisely preserved across the entire conformation or oligomerization ensemble, in which case anesthetic binding will not alter the distribution of conformers and therefore will have no effect on activity. How commonly this form of "unproductive" binding occurs is not clear.

Summary

Any or all of these general coupling mechanisms may be involved in anesthetic-induced protein dysfunction. The distinction often is difficult but is important if we are to intelligently modify the drugs to favor or disfavor specific interactions.

Potential Protein Targets

As suggested earlier, the anesthetic sensitivity of a number of protein systems has been studied over the past few decades, and, remarkably, many are altered within only a 10-fold range of clinical concentrations. We briefly review some of these systems and the evidence for inclusion. This discussion is not intended to be a comprehensive list of those proteins studied; many have come and gone. Rather, we focus on some of the more recent and compelling candidates, especially those with associated in vivo evidence for a contribution. Furthermore, we focus on targets associated with the desirable effects, such as hypnosis, rather than the many that might underlie less desirable side effects.

Soluble

As pointed out earlier, proteins have domains as hydrophobic as those of the lipid bilayer, and it now is clear that anesthetics bind to and influence the activity of many soluble proteins. Much of the published work involved model systems, such as firefly luciferase, serum albumin, and apoferritin, which are unlikely to contribute directly to anesthesia. However, plausible candidates, centrally located in signal transduction cascades or cellular machinery, have emerged. For example, protein kinase C (PKC) transduces receptor activation to phosphorylation of key membrane proteins, such as certain ion channels and other receptors, and has been shown to bind general anesthetics.31 Alteration of its activity by anesthetics would be expected to have generalized and widespread cellular effects that might contribute to anesthesia, although controversy exists as to both the expected and observed directions of effect (reviewed by Rebecchi and Pentyala32 and Gomez and Guatimosim33). Initially thought to be inhibited by general anesthetics, later studies performed under physiologic conditions showed PKC to be activated. This one system demonstrates the complexity of sorting out anesthetic targets and effects. The observed effects in this relatively simple physiologic assay can result from anesthetic effects on membrane lipid, stimulus–kinase coupling, kinase–target coupling, or the phosphorylated target itself. Anesthetics might interact with any component. Studies in intact, genetically altered animals are in progress but to date have not provided unambiguous answers. Further studies on PKC isoform distribution and activity are required to sort out the influence of anesthetics and potential contribution to the anesthetic state.

Another attractive anesthetic target is any of the components of the cellular cytoskeleton, motility apparatus, and vesicle transport systems. These systems subserve a variety of basic cellular functions, the alteration of which would manifest as a decrease in cellular activity or communication. For example, tubulin forms microtubules, which are crucial elements of cellular scaffolding and motility. Anesthetics bind specifically to tubulin,34 and the assembly of the monomer into the large tubular oligomer is altered by anesthetics, albeit at high concentration.35 Entire mechanisms for general anesthesia based on microtubular dynamics and organization have been proposed,36 and some experimental support has been reported.37 Oligomerization of actin, a protein acting with myosin and other proteins to subserve cellular motion, is inhibited by anesthetics and is another relatively unexplored potential general mechanism of anesthesia.38

Anesthetics also affect the synaptic vesicle machinery. A decrease in synaptic transmission is generally agreed to accompany general anesthesia; therefore, inhibition of synaptic vesicle transport, fusion, and release is a plausible general mechanism of central nervous system (CNS) dysfunction. Evidence comes from several disparate sources. In a genetic screen of the nematode Caenorhabditis elegans, mutations in the vesicle fusion soluble N-ethylmaleimide sensitive attachment factor (SNARE) proteins were found to produce resistance to some anesthetic endpoints.29 Furthermore, 1 component, syntaxin, and various SNARE assemblies have been shown to bind isoflurane using nuclear magnetic resonance approaches.39 Transcript profiling40 in mammals has shown an upregulation of synaptotagmin, another vesicle-release protein. The latter does not necessarily implicate synaptotagmin as a direct anesthetic target, but it lends credence to the idea that synaptic vesicle release is inhibited during general anesthesia. Other synaptic targets implicated are the PDZ domain proteins. These domains are responsible for fusion events (oligomerization) between proteins and are intimately involved in synaptic vesicle trafficking. Work has shown that anesthetics disrupt the oligomerization, and the probable anesthetic binding site responsible has been identified in a truncated version of PSD-95.28

Membrane

Ion Channels

A variety of proteins embedded in, and dependent on, the lipid bilayer for their activity are altered by general anesthetics in in vitro assays. Most studied are the ion channels, largely because of the availability of electrophysiologic approaches for measurement of function. Of the ion channels, most emphasis has been placed on the ligand-gated cys-loop receptor–channel complex, the prototype of which is the nicotinic acetylcholine receptor (nAChR). These proteins are transmembrane heterooligomers of 5 subunits arranged around a central ion channel. A variety of anesthetics inhibit this excitatory channel, perhaps by promoting the desensitized state via cooperative binding with acetylcholine; the probable basis for the muscle relaxation associated with many general anesthetics. Volatile anesthetics bind this receptor specifically, and a site underlying cooperative binding behavior has been tentatively identified.20,22 The nAChRs are also found in the brain, adding credibility to the possibility that alterations in their activity contribute to anesthesia. In vivo support of an important role for these receptors is still lacking.

γ-Aminobutyric acid (GABA)ergic neurotransmission was identified 20 years ago as a plausible substrate for anesthetic effects,41 so recent emphasis has been placed on the inhibitory GABA type A (GABAA) and glycine receptor–ion channel complex as potential contributors to at least the sedative or amnestic component of general anesthesia. Similar to the nAChR, many general anesthetics (most volatile ones and alcohols) probably bind cooperatively with agonist (either GABA or glycine). However, because these receptors undergo desensitization more slowly that the nAChR, the final effect is enhanced agonist-stimulated currents instead of inhibition (Fig. 37-14). These ion channels are inhibitory (produce hyperpolarization of neuronal membrane by opening of chloride channels), so enhancement of their activity is expected to inhibit synaptic transmission. The difficulty of isolating or expressing sufficient GABAA receptor of any subunit composition for biochemical studies so far has precluded the demonstration of specific volatile anesthetic binding, although an etomidate site has been identified at an intersubunit interface.42

Most studies to date have focused on the action of volatile anesthetics on GABAA receptors mediating rapid synaptic transmission. However, volatile anesthetics also act on GABAA channels mediating tonic inhibition. Tonic GABAA channels are found in many areas of the nervous system, including the hippocampus and the spinal cord interneurons. Although the precise receptor subunits composing the tonic GABAA channels remain unknown, the α5- and δ-subunits likely are involved because tonic currents inhibited by bicuculline are not seen in null-mutant mice devoid of these subunits. These GABAA channels are more sensitive to agonists than the rapid type, requiring only low micromolar concentrations of GABA for opening, concentrations found in even extrasynaptic regions of a nerve cell. Similarly, tonic GABAA channels are more sensitive to volatile anesthetics than those mediating rapid synaptic transmission.43 Because of this sensitivity and expression in the hippocampus, it is reasonable to hypothesize that these channels mediate the amnestic component of volatile anesthetic action. A thorough characterization of the volatile anesthetic effects on the various tonic GABAA null mutant mice has not been reported.

In an attempt to both find sites of action as well as to implicate anesthetic targets, site-directed mutagenesis has been coupled with electrophysiology. For example, an asparagine to methionine mutation at position 265 in the GABAA β3 subunit largely eliminated enflurane enhancement of GABA action in vitro, but the knock-in mouse, in which the wild-type β3-subunit was replaced by the point mutant, had minimally altered righting reflex sensitivity or MAC.44,45 Greater success was obtained with the injectable general anesthetics (induction agents) in this same model. For example, similar to enflurane, etomidate and propofol failed to enhance GABA-evoked currents in vitro, but in this case, the whole-animal sensitivity to both injectable drugs was significantly diminished. The technology of mutant receptor knock-in is far superior to an overall elimination of the receptor produced by the knock-out technology because the normal cellular process that regulates expression of the wild-type receptor is preserved. Based on structure–activity studies demonstrating the role of both α- and β-GABAA subunits46 as well as other ligand-gated channels in determining volatile anesthetic responsiveness in in vitro expressed receptors, a multiple knock-in mutant mouse in which many sites are simultaneously altered is more likely to exhibit relative resistance to the volatile anesthetics.

Interestingly, in contrast to prototypical volatile anesthetics such as isoflurane and injectable anesthetics such as propofol, etomidate and barbiturates, a group of agents called dissociative anesthetics, similar to ketamine and nitrous oxide (N2O),47 as well as xenon,48 at clinically relevant concentrations, do not significantly affect GABAA currents, but they inhibit a major excitatory drive in the CNS via blockade of N-methyl-D-aspartate receptors (NMDAr). Thus, when ligand-gated ion channels are concerned, general anesthetics may inhibit CNS neurons either by potentiating inhibitory GABAA currents or inhibiting excitatory NMDAr-mediated currents. Because NMDA receptors play a key role in pain pathways in CNS, their blockade may contribute to the prominent analgesic properties of dissociative anesthetics.

Another hypothesis is that the 2-pore domain family of background K (KCNK) channels is involved in volatile anesthetic action. This gene family encodes for K+-selective ion channels with a common structural feature of 2 protein domains putatively lining the ion channel and 4 transmembrane domains. Two 2-pore domain proteins are thought to assemble as a dimer, creating a complete protein complex with a functional K+-permeable ion channel normally open at all physiologic membrane potentials, thus contributing to the background leak K channel critical in determining the resting membrane potential and regulating neuronal excitability.49 Opening of channels encoded by 2 members of the KCNK family, TREK-1 and TREK-2, and more recently a third member, TRESK, are enhanced by volatile anesthetics, thereby hyperpolarizing the cell membrane. However, the enhancement is not seen in all the family TRESK because the channel activity of a closely related member, TRAAK, is not affected. As in the case for GABAA receptors, the hypothesis that the anesthetic-responsive KCNK channels play a role in the general anesthetic action at the whole animal level was examined and confirmed in TREK-1 knock-out mice. For example, mice without the TREK-1 gene were approximately 20% less sensitive to halothane, sevoflurane, desflurane, and chloroform (but not pentobarbital) as defined by the loss of righting reflex (LORR) and withdraw to tail clamp. Interestingly, an invertebrate analog of this channel41 was discovered a decade earlier in the pond snail,49 pointing out the importance of information derived regardless of the model organism used. Voltage-gated potassium channels are generally insensitive to anesthetics, although a couple variants found in Drosophila and in mammalian orthologs are sensitive to clinical concentrations of both alcohols and anesthetics.50

Voltage-gated calcium channels (VGCCs) have been studied for years because of the central role of calcium in intracellular signaling. VGCCs, which are heteromeric complexes in the plasma membrane of virtually all cell types, show a high level of electrophysiologic and pharmacologic diversity. These channels consist of a pore-forming α1-subunit and ancillary subunits β, γ, and α2–δ.51 On the basis of the membrane potential at which they activate, these channels are subdivided into high voltage-activated (HVA) and low voltage-activated (LVA) or transient T-type Ca2+ channels (T channels). These channels in nerve tissue have a central function in sensory, cognitive, and motor pathways and controlling cell excitability and neurotransmitter release. These channels, which are products of different genes, give rise to α1 subunits that form the pore of the neuronal VGCCs. The HVA VGCCs are members of different families: CaV1 (α1C) encoding L-type, CaV2.1 (α1A) encoding P/Q-type, CaV2.2 (α1B) encoding N-type, and CaV2.3 (α1E) encoding R-type HVA current. Similarly, cloning of T-type channels has established that at least 3 isoforms exist based on the structure of α1 subunits: CaV3.1 (α1G), CaV3.2 (α1H), and CaV3.3 (α1I). Because of its importance, intracellular calcium is tightly controlled via many systems, including voltage- and ligand-gated channels, exchangers, ATPases, and soluble binding proteins. Many of these systems have been shown to be influenced by anesthetics, although early studies have shown that, in general, the magnitude of effect is modest at clinical concentrations.33 However, in most instances, previous studies used only 1 general anesthetic; did not pharmacologically separate subtypes of VGCCs in native cells; and in many cases, did not obtain careful concentration-response curves. An important issue is that even small blockade of a particular VGCC by any general anesthetic may produce profound physiologic effects. When this has been examined, small changes in Ca2+ influx into presynaptic terminals can result in profound changes in transmitter release and synaptic efficacy.51 Indeed, presynaptic transmitter release is proportional up to the fourth power of Ca2+ entry.51 Thus, even if VGCCs are only partially inhibited by anesthetics in the clinically relevant range, this can profoundly alter neuronal signaling. The effects of volatile anesthetics on even 1 subset of VGCCs are variable, probably reflecting the molecular heterogeneity of these channels. Of the functionally characterized VGCCs, the N- and P/Q-type channels, which are major putative presynaptic channels regulating neurotransmitter release, were only partially inhibited by clinically relevant concentrations of inhaled anesthetics.52,53 The recent evidence supporting a possible role of inhibition of N-type VGCCs in the pharmacology of inhaled anesthetics comes from the observation that N-type VGCCs knock-out mice exhibited increased sensitivity to halothane,54 although as noted elsewhere in this chapter, this is an extremely common observation regardless of the target manipulated. Another recent study using in vitro recordings from intact thalamic slices and in vivo recordings using CaV2.3 R-type knock-out mice has established that presynaptic R-type VGCCs in the thalamus are potently inhibited by clinically relevant concentrations of isoflurane.55 The T-type VGCCs channels, although not involved in synaptic neurotransmitter release, play an important role in controlling neuronal excitability and in generating spontaneous oscillatory bursting of groups of neurons in the thalamus thought to be involved in regulating the state of arousal and sleep. The thalamic T-type channels are significantly blocked by clinically relevant concentrations of the inhaled anesthetics.56 The molecular mechanism of how these drugs inhibit the VGCCs channels is not known but appears to involve acceleration of channel inactivation and slowing of the recovery from this nonconducting state.56 There is no evidence of anesthetic binding to these targets, so whether this electrophysiologic effect is mediated via anesthetic interactions with lipid or protein is uncertain.

Voltage-gated sodium channels (NaVs) play a key role in regulating neuronal excitability and therefore are a plausible target for the inhaled volatile anesthetics. However, NaVs were largely dismissed as a relevant target because initial studies indicated that high concentrations were required for inhibition in squid axons (reviewed by Elliott et al57). However, later studies of mammalian brain NaVs indicated that these channels are significantly inhibited by lower, more clinically relevant concentrations of volatile anesthetics.58 In fact, these anesthetics influence specific ligand binding to GABAA receptors and the NaVs with approximately equal potency,59 and rank-order effects appear to qualify the presynaptic NaV as a feasible target for contributing to general anesthesia (Box 37-1).

Receptors

The dominant receptor type in the brain is the GPCR. This enormous family includes most of the neurotransmitter and sensory receptors, whether small molecule, peptide, or lipid. These receptors are monomeric and have 7 transmembrane domains arranged like an envelope around a central cleft, or cavity. On the cytoplasmic face, GPCRs couple to the heterotrimeric (3 different subunits) G-protein messenger systems. The native ligand binding site tends to be at various depths in the interhelical cavity, and although some features are conserved, the site can accommodate a wide variety of ligand structure and chemistry (Fig. 37-15). Ligands range from small molecules (volatile odorants, catecholamines) to small peptides (endogenous opioids). That anesthetics act on these receptors is suggested by the anesthetic-sparing effect of agonists for the dopamine and α2-adrenergic receptors in addition to alterations of activity of the eicosanoid receptors.61 Finally, volatile anesthetics have been shown to bind the conserved agonist site in many of these receptors (Fig. 37-15),24 producing either inhibition or excitation.62 Effects downstream of these receptors (eg, on the G-protein transduction pathway) are also possible. Evidence is ample for anesthetic effects on G-protein–mediated signaling32,61 and the importance to function in vivo. For example, C. elegans with diminished Go activity (a negative modulator of synaptic activity) was 2-fold resistant to isoflurane.63 These effects probably are related to effects on the receptor, or the receptor–G interface, because evidence for direct binding to any G-protein subunit or the heterotrimeric complex has not yet emerged.32 This work is highly intricate because of the multiple complexes and states and at this time is unresolved.

Mitochondria

Mitochondria have long been suspected as contributing to anesthetic action because of their obvious role in cellular energy production. However, the anesthetic concentrations required for inhibition of ATP production and the apparently slow kinetics for ATP depletion seemed to obviate their role. Recent evidence has renewed interest in the mitochondrion. First, signaling and feedback between mitochondria and cellular consumers of ATP is considerably more precise and rapid than originally thought, effectively eliminating the kinetic argument against their contribution to the anesthetic state. Second, mitochondrial genes have been implicated in unbiased genetic screens in simple organisms (see later discussion). For example, mutations in a mitochondrial complex I subunit gene dramatically increased sensitivity to halothane, enflurane, and isoflurane. As expected, mitochondria isolated from this mutant had a reduced rate of oxidative phosphorylation in the presence of halothane.64 Furthermore, several subunits of the oxidative phosphorylation complexes, including complex I, were found to bind halothane specifically.65 Perhaps most importantly, children with biopsy-proven complex I disease were found to be extraordinarily sensitive to the anesthetic sevoflurane,66 rendering this anesthetic target the only to date that has supportive evidence from the gene, the binding interaction, to the human. The underlying mechanism for how the altered mitochondrial function affects anesthetic sensitivity is unclear, but given the recent evidence that mitochondria modulates synaptic transmission through the regulation of presynaptic Ca2+ dynamics,67 it is possible that previously described mutations affect synaptic transmission. Another potential mitochondrial target is the voltage-dependent anion channel 1 (VDAC-1), an anion channel also related to mitochondrial steroid synthesis and synaptic activity and implicated through halothane and neurosteroid binding assays, although knock-out animals had unaltered anesthetic sensitivity.65,68,69

Summary

A wide variety of molecular candidates have been examined, and although several are compelling, proof for dominant involvement in the in vivo anesthesia endpoint is still lacking. But how does one determine the relevance of a molecular interaction to the in vivo effect, especially when it is as complex as consciousness?

How to Determine the Relevance of In Vitro Assays

The plethora of potential targets, the imprecise definitions of anesthesia, the variable endpoints, and the low-affinity drugs have conspired to make it difficult to judge the contribution of molecular targets to drug behavior, an already difficult task. As a start, 2 broad concepts of anesthetic action have been proposed: one favors a single molecular target site (the "unitary" hypothesis), and the other proposes that anesthesia results from the combination of actions at many molecular targets (distributed hypothesis). Lipid theories are an example of unitary theories, as are those that focus solely on the GABAA receptor. Distributed action hypotheses require careful dissection of more molecular targets. Historically, the hunt for the volatile anesthetic targets has taken 2 general approaches: the study of effects on a well-defined and functionally or physically isolated target (bottom-up approach) and those that start with the anesthetic effect on an organism with the goal of defining the target responsible for the observed effect using either genetic or pharmacologic probes (top-down approach). Although both approaches have merits and limitations, it is the complex, dynamic coupling between "top" and "bottom" that has so far made clear conclusions difficult. Next we discuss how information derived from one approach might contribute to testable hypotheses in the other and how when taken together, support for a neurophysiologic process (the "middle") rather than an individual protein or behavior emerges as a likely anesthetic target. It is important to realize that disruption of a process might have multiple contributing and interacting molecular events.

"Bottom-Up" Approach

This approach has dominated much of the literature on the mechanism of general anesthetic action. Examples are studies of the biochemistry of anesthetic action on a specific signaling cascade, such as activation of a G-protein coupled pathway, studies on the structural nature of anesthetic binding sites in synthetic proteins, and electrophysiology of heterologously expressed ion channels. An important weakness of this approach is that it requires the identification of a system for study based on the inherently biased criterion, plausibility.

Plausibility has driven studies on the effects of anesthetics on the electrophysiology of ion channels expressed in heterologous systems such as the Xenopus oocyte. Such electropharmacologic studies have demonstrated effects (enhancement or inhibition) of anesthetics on essentially all the ion channels examined, albeit some requiring higher concentrations of anesthetics than others. Herein lies the principal weakness of the bottom-up approach. How do we determine whether a given effect on a given target contributes to a particular behavioral endpoint? Discrepancy between the anesthetic concentration required for the effect on a given target protein and the clinical MAC (either positive or negative) has led to a call for dismissal of some putative targets as biologically irrelevant.70 As we discussed for the relationship between Kd and EC50, the relationship between EC50 and MAC is not likely to be linear. The nervous system is complex, and inclusion of even 1 simple nonlinear process, such as generation of an action potential in a synaptic transmission (a threshold effect), can easily shift an apparently irrelevant in vitro concentration–response relationship to one that is quantitatively consistent with clinical MAC (Fig. 37-16).51 Even at the single synapse level, spatial and temporal integration of postsynaptic currents contributing to the generation of action potentials will interpose further nonlinear processes between a drug action at a receptor and an observable output. A simple cascading of multiple processes, each well described by a conventional continuous concentration–response curve (ie, Hill equation), will further shift the composite concentration–response curve to the left with an increase in the Hill slope.71,72 Thus, a quantitative discrepancy in the experimentally and clinically observed concentration–response relationship cannot discredit the biologic relevance of a putative anesthetic target.

Given the apparent difficulty of linking molecular events and behavior, other "litmus test" criteria have emerged. For a target to be seriously considered as a candidate, most agree that the anesthetic must actually occupy a binding site in or on the target (see earlier). Thus, it is remarkable how many candidates are being favored in the complete absence of binding data. Other criteria used to establish relevance are Overton-Meyer behavior (correlation of potency with lipid solubility), enantioselectivity, sensitivity (correlation of in vitro EC50 with MAC), lack of response to nonimmobilizer compounds, hydrostatic pressure reversal, cutoff, and plausibility. For example, if differences in potency exist in the intact organism for the enantiomers of a given anesthetic, then the targets responsible for this anesthetic effect also should exhibit this difference in potency in vitro. If the organism fails to lose consciousness with a particular compound (eg, F6 or F8), then any underlying molecular target also should remain unaffected.73 However, this litmus test logic assumes that a single target can produce anesthesia, and now that several molecular targets have emerged that satisfy many of these criteria, such criteria should be viewed with considerable skepticism.

A final means of establishing linkage of the molecular event with the behavior involves use of specific pharmacologic probes. For example, concentration–effect experiments with anesthetics can be performed in the presence of other sedative drugs, and then the interaction can be characterized as synergistic, additive, or antagonistic through isobolographic analysis. If the second drug is known to be highly specific for a given molecular target, then some insight might be gained into the targets used by anesthetic. Although this approach has identified many potential targets for the inhaled anesthetics, its use has been somewhat reduced by the fact that the second drug rarely produces the same endpoint as that of the anesthetics. For example, the opioids and the benzodiazepines, unless given in very high doses, do not produce the MAC endpoint, but this is the standard endpoint for the volatile anesthetics. Yet in both cases, the second drug decreases the volatile anesthetic MAC by 10% to 50%. Agonism at the GPCR α2-adrenergic receptors reduces anesthetic requirements by as much as 90%,74 although whether synergistic or additive cannot be clear because these agents cannot independently reach this endpoint. Finally, this pharmacologic approach is further confounded by the fact that the effects are global, and it is now recognized that the same targets may have important influences only regionally (see later).

Summary

Bottom-up studies allow for well-controlled experiments with well-defined endpoints (eg, examination of the extent of anesthetic potentiation of GABAA receptors) and for straightforward subsequent genetic manipulation of the target protein through conventional site-directed mutagenesis of the complementary DNA encoding the protein of interest expressed in vitro. However these approaches are primarily useful for generating testable hypotheses, in which the biologic relevance of the putative target can be proposed and then examined in a more intact system, such as with targeted pharmacologic or knock-out, knock-in, or knock-down genetic approaches.

"Top-Down" Approach

An alternative approach to determining the sites and mechanisms of drug action is a "top-down approach" in which the starting point is an individual or a population of intact organisms. Populations typically harbor mutations in their genome, and thus individuals with high or low sensitivity to volatile anesthetics can be identified through standard potency experiments and the potential molecular basis hypothesized based on the discovered genetic differences. For example, strain differences in mice have been demonstrated75 but are small (<50%), and the underlying genetics appear to be complex. This approach is limited, however, because it requires that considerable variability exist within the population of organisms to clearly select for the differences. The Hill slopes of 20 in the human population suggest that sufficient variability in the response does not exist for this to be a productive approach unless enormous numbers of individuals can be tested.

Although difficult in human studies, such numbers can be easily studied in simple organisms, such as the fruit fly Drosophila melanogaster or the worm C. elegans. These systems give the added advantage of relatively easy identification of the genetic defect from the selected population, an approach called "forward genetic" screening. The primary advantage of such a system lies in the possibility of revealing a direct correlation between a genetic target and a complex behavior involving no a priori assumptions or biases. Furthermore, the natural prevalence of mutations often can be considerably enhanced by chemicals or ultraviolet light in these simple systems, greatly increasing the odds of finding interesting alterations. On the other hand, the fundamental limitation of simple model systems lies in the definition of the anesthetic state or the behavioral endpoint used and its relevance to the state of general anesthesia in higher organisms. For example, the yeast Saccharomyces cerevisia is an enormously useful eukaryote for genetic studies, but how can one possibly define anesthesia in the absence of behavior? Is anesthetic-induced immobility in a nematode analogous to anesthesia? Should similar endpoints be chosen across species (eg, immobility) independent of the anesthetic concentration dependence, or should arbitrary endpoints be chosen that happen to have similar concentration–response relationships across species? Will determining the genetic basis for anesthetic-induced growth inhibition in yeast or escape behavior in flies tell us anything about how anesthetics work in humans? Although such questions are legitimate and vexing, conservation of biologic processes throughout phylogeny has been proven time and again, and a fundamental biologic mechanism operating in simple organisms is likely to have a homologous mechanism in the more complex organisms. Research using the forward genetic approach to seek the volatile anesthetic site of action in simple organisms has yielded several exciting potential targets.

Similar to humans, nematodes have different behaviors that demonstrate different sensitivity to anesthetics. For example, whereas complex behaviors such as mating, chemotaxis, and coordinated movement are inhibited at halothane concentrations similar to human MAC values,29 immobility occurs only at concentrations approximately 8 times higher (Fig. 37-17).76 The top-down screening of a de novo mutagenized population using immobility as the endpoint has identified genes encoding the worm homologues of 2 mammalian proteins as the genetic basis for increased sensitivity to anesthetics (resistant organisms for the immobility endpoint have not been discovered): a subunit of mitochondrial complex I protein (see earlier) and stomatin, a transmembrane protein enriched in the lipid raft microdomains. Interestingly, these studies have clearly documented that these mutations do not affect sensitivity to all volatile anesthetics, the first solid evidence against a unitary hypothesis of volatile anesthetic action. A very powerful tool in these simple creatures is that suppressors (ie, strains with additional mutations that negate the phenotype of the original mutation) can be sought. Identification and function of the suppressor then can be reconciled with that of the original mutation to provide validation and further understanding of the underlying pathways and mechanisms. For example, a suppressor of the worm mitochondrial complex I mutation was found that increased the capacity for oxidative phosphorylation in isolated mitochondria and partially restored-wild type anesthetic sensitivity in the intact worm.77 This, when combined with the observation that children with complex I disorders have unusual high anesthetic sensitivity,66 strengthens the association between anesthetic sensitivity and mitochondrial function.

A similar screening of C. elegans mutants using the loss of more complex behaviors, such as coordinated movement, as the behavioral endpoint identified the syntaxin-1A gene, mentioned earlier,29 as being associated with anesthetic resistance (≤ 6-fold). Further work, including binding studies and suppressors, has provided evidence for the involvement of syntaxin-1A in anesthetic action and has led to the attractive and intuitive hypothesis that anesthetics interfere with the release of neurotransmitters at the presynaptic terminal, producing a general reduction in synaptic transmission (see later).78 Again, these simple creatures have provided compelling hypotheses for testing in higher organisms.

Moving up the complexity and evolutionary scale, studies in D. melanogaster have documented that specific ion channels can contribute to volatile anesthetic sensitivity (Fig. 37-18). For example, forward genetic screening has succeeded in isolating a putative ion channel resembling the voltage-gated sodium and calcium channels as important for anesthetic-induced immobility.79 As in the worm, the different ion channels show variable sensitivity to different volatile anesthetics, suggesting that multiple neurophysiologic mechanisms mediate the action of different anesthetics even in simple organisms.

A final related approach that avoids investigator bias is assaying for the genes or proteins whose expression is altered during or after exposure to volatile anesthetics. This approach carries the tenuous assumption that the cells and organism will choose to alter the expression of those targets directly affected by the anesthetic in an attempt at homeostasis. In some published studies, the expression of many genes or proteins is altered.80,81 However, using a more (perhaps overly) rigorous statistical approach, a study found only a modest transcriptional response after considerable exposure to either halothane or isoflurane in both animals and cells.40 It is still too early to summarize this approach because the methodology, analysis, and interpretation are evolving.

Taken together with the evidence presented, observations from both bottom-up and top-down studies suggest that a search for a small number of targets for volatile anesthetic action in a complex organism is overly simplistic and unlikely to yield a satisfactory or interpretable answer. However, future studies using novel methods for confirming a critical role of a given genetic defect in determining the inhaled anesthetic sensitivity may yet prove that our arguments against the unitary hypothesis are incorrect.

Integration of Molecular Responses

If anesthesia is an integrated response to a large number of molecular events, then validation at the molecular level or in translational attempts of single candidates will be difficult. For example, the most popular means of testing hypotheses from both bottom-up and top-down approaches is to produce genetically altered animals (usually mice). As mentioned for GABAA receptors, this form of validation suffers from at least 2 weaknesses. First, in a knock-out (null mutant) animal, developmental compensation may occur in multiple associated pathways, possibly providing new mechanisms for anesthetic-induced perturbations. However, this does argue against highly specific "unitary" targets because the likelihood that such adaptation in other components of the pathway will produce similarly unique and functionally active binding sites seems rather small. The second weakness, for both knock-out and knock-in strategies, is that the physiologic alteration may be of sufficient magnitude to render endpoint evaluation difficult (eg, seizures, lethality). Consistent with these concerns, this means of testing individual targets has generally yielded ambiguous results.

The existence of a large number of targets does not fully rule out a unitary mechanism of action, however. Such a mechanism might be found at a higher level of resolution, such as at the level of a neurophysiologic process. Candidates at this level are numerous but could include transcription or translation, mitochondrial activity, membrane resting potential, action potential conduction, and synaptic transmission, ultimately leading to changes in the output of neuronal circuits. Each process has multiple and overlapping molecular contributors.

Decades of study have implicated the synapse as a key neurophysiologic substrate affected by volatile anesthetics.82 This idea initially was based on the simple observation that axonal conduction of action potentials was more resistant to volatile anesthetics than synaptic transmission. Insight gained from the large number of bottom-up and top-down studies of volatile anesthetics has supported and refined the notion that synaptic transmission is the most likely neurophysiologic process targeted by these drugs (Fig. 37-17).

Synaptic transmission, although the most fundamental unit of intercellular information transfer in the nervous system, is a highly complex process in itself (Fig. 37-19). When an action potential arrives at the presynaptic terminal, an increase in presynaptic Ca2+ concentration via voltage-gated calcium channels and release from intracellular stores activates the synaptic vesicle release machinery orchestrated by and consisting of a myriad of proteins. After release into the synaptic cleft, the concentration of transmitter available to the postsynaptic receptor is influenced by neurotransmitter uptake pumps and degradative enzymes for some synapses. On activation of the postsynaptic receptor (G-protein coupled or ligand-gated channel) and the consequent influx of ions and activation of cellular signaling molecules, the change in charge distribution may produce depolarization of the postsynaptic membrane and formation of an action potential if the threshold is exceeded. Thus, it should be clear that anesthetics can and do have an effect at each of the many points in the process.

Given a focus on synaptic transmission as a key process targeted by volatile anesthetics, some predictions can be made on the anesthetic phenotype of some mice not yet examined. For example, the null-mutant mice for the various calcium channels or release mechanisms that regulate presynaptic calcium dynamics, all with impaired synaptic transmission, should demonstrate increased sensitivity to volatile anesthetics. In contrast, mouse mutants with enhanced synaptic transmission, which may result from mutations increasing neurotransmitter concentration in the synapse, should exhibit decreased sensitivity to anesthetics. If we assume that the state of clinical general anesthesia is a manifestation of all the downstream consequences of the failure of synaptic transmission, there will be many more yet unidentified target proteins and processes contributing to this complex drug action. However, interpretation of the anesthetic phenotype of any gene-targeted mice must take into consideration the likely redundancy built into any complex system and the possibility of compensatory changes in the conventional null-mutant mice.

A model in which a process is altered by widely distributed and perhaps small input effects also predicts difficulty in altering the output with isolated input manipulations. Analogy may be found on the Internet, where many coordinated small and essentially undetectable influences on router function (colluding routers) can have a considerably larger effect on Internet "function" than a few large and easily detected router "meltdowns."86 A small-effects-at-multiple-targets model leads to predictions that so far have been largely verified. For example, altered plausible molecular targets with clear alteration of anesthetic effects in isolation, when introduced in a system or process context, produce little, if any, effect on anesthetic sensitivity. Although this is complicated in biologic systems by homeostasis and compensation, this observation appears to be consistent. In no case has a mutation of an individual target or highly specific drug altered anesthetic potency by more than 2-fold, usually much less. Furthermore, the facts that antagonists have not emerged, that there is little biologic variability, and that there is no adaptation even when the drugs are given continuously for weeks all argue for a highly distributed, redundant molecular effect. Redundancy is often used in systems design to produce reliability, a feature synonymous with low variation. The reliability of anesthetic action is a clinical feature we cannot afford to ignore, but its foundation in a distributed mechanism is a scientific feature that will be difficult to sort out.

Neuroanatomic Substrates

A process may be widely distributed within the nervous system (eg, synaptic transmission), but the influence of its dysfunction could be sensed and exhibited regionally. Although there is no evidence for regional binding of anesthetic (Fig. 37-8), binding targets may be positioned within networks and pathways that rely on their neurotransmitter types and regional interactions for higher functions, such as behavior. Thus, GABAergic enhancement in the spinal cord or the cerebellum may have very different functional effects than enhancement in the hippocampus. The different functional components of general anesthesia likely arise from such regionality (Box 37-2).

Thus, volatile anesthetics produce a composite state that can be subdivided into functional components, which include amnesia, unconsciousness, analgesia, muscle atonia, and autonomic nervous system modulation. Each may represent specific interactions of general anesthetics upon discrete neuronal loci. Just as the search for molecular targets of anesthetic action has revealed multiple sites of modulation, so too has the hunt for cellular sites of anesthetic action. In this section, we focus upon different features of the anesthetic state that arise from anesthetic actions in various parts of the CNS.

Anesthetics and Immobility

The MAC endpoint is defined as the MAC of an agent that is required to suppress movement in response to a defined noxious stimulus. It should not be surprising then that anesthetics appear to exert their immobilizing effects most sensitively through interactions within the spinal cord. For example, transection of the brain at the level of the inferior colliculus does not alter the MAC of inhaled anesthetics in mammals.87 Although such lesions ablate ascending and descending forebrain and midbrain communications with the spinal cord, they do not impair spontaneous ventilatory efforts because brainstem respiratory groups are left intact. This works suggests that MAC is primarily determined by a direct action of anesthetics upon the CNS at a level caudal to the brainstem—the spinal cord. Further work in multiple species supports this hypothesis. Surgical isolation of brain and spinal cord blood flow in goats (which lack significant cerebral perfusion from vertebral arteries) allowed independent determination of MAC. Delivery to the brain required a greater than 2-fold higher concentration than did whole-animal MAC, suggesting that immobility is produced via direct interactions within the spinal cord.88,89 In vitro evidence of anesthetic effects on the spinal cord has also been obtained. All volatile anesthetics tested to date appear to suppress the excitability of spinal motor neurons; however, even within the spinal cord, different anesthetics exert their effects at multiple sites. For example, whereas halothane reduces noxious-evoked movement via depression of dorsal horn neurons, isoflurane appears to have smaller effects on these sensory neurons yet produces an identical inhibition of movement by exerting a relatively greater suppression of anterior horn motor neurons or interneurons.90

Anesthetics and Hypnosis

Although the neural substrate responsible for generating consciousness remains unknown, much has been learned about the systems responsible for its nightly loss and subsequent return during the normal process of sleep. The physiologic changes associated with unconsciousness induced by both anesthesia and sleep are strikingly similar. They include decreased minute ventilation, blood pressure, heart rate, core body temperature, motor tone, and (most importantly) responsiveness to external stimuli. Studies of cortical function in both states reveal a similar breakdown of spatial and temporal information transfer.91,92 The slow-wave oscillations in the electroencephalogram (EEG) that define the deepest stage of non–rapid eye movement (NREM) sleep bear striking resemblance to those seen under hypnotic doses of some anesthetics. In both instances, slow waves originate in the same "hotspots" and propagate in similar directions throughout the cortex.93 Moreover, the spindle waves common to NREM sleep are also seen during general anesthesia.94 Positron emission tomography, used to study cerebral blood flow during both sleep and anesthesia, has confirmed that similar alterations in regional cerebral blood flow, metabolism, and regional neuronal activity occur during both sleep and general anesthesia.95 These similarities and others (documented in the following) have led to the speculation that sleep and general anesthesia share an underlying neural substrate. Consistent with this hypothesis is the observation that an individual's underlying state of arousal alters anesthetic sensitivity. In rodents, loss of consciousness typically is assessed by the LORR behavioral assay in which animals turned from supine to prone are unable to right themselves. Using this endpoint, Einon et al96 noted that the hypnotic duration of pentobarbital was prolonged when the drug was administered to rats during their sleeping hours compared with their normally wakeful hours. Sleep deprivation acts synergistically with anesthetics, reducing the dose of anesthetic required for hypnosis and allowing recovery from sleep deprivation.97

Further linkage between sleep and anesthesia derives from pharmacologic studies. Endogenous somnogens such as adenosine, which can induce sleep, also potentiate the hypnotic effects of inhaled and intravenous anesthetics and delay emergence from anesthesia.98 Moreover, anesthetic exposure appears to affect levels of endogenous somnogens, such as prostaglandin D2.99

Despite these overlapping features, it is clear that sleep and general anesthesia have important differences. The most notable is that whereas sleep is readily reversed through external stimuli, anesthesia is reversed only upon discontinuation of the anesthetic drugs. Another primary difference is that anesthetics appear to inhibit systems required for rapid eye movement (REM) sleep as well as cortical arousal.94 Hence, general anesthesia has been likened to a form of NREM sleep from which one cannot be easily aroused.95

The similarities between NREM sleep and general anesthesia extend to various measures of the EEG. For example, both power spectrum analysis and the bispectral index100 show a high degree of similarity between NREM sleep and general anesthesia. EEG entropy, a measure of the disorder inherent in the EEG, declines during deep NREM sleep and during general anesthesia.101,102 In both instances, as NREM sleep or anesthetic depth increases, the EEG becomes more ordered, further reducing EEG entropy.

Neural Circuitry of Arousal

Similarities between general anesthesia and NREM sleep have led to the speculation that the 2 states share a common underlying neural network. As is the case with the molecular targets, there are many potential neuronal structures upon which anesthetics can act to produce unconsciousness. Because of bidirectional signaling (a neuron may send efferent fibers to and receive afferent fibers from its signaling partner), it has been difficult to unravel a primary site of anesthetic-induced hypnosis. Nonetheless, putative sites of action for the hypnotic effects of anesthetics are found along the entire neuroaxis from the brainstem through the thalamus, hypothalamus, basal forebrain, and cerebral cortex (Fig. 37-20). Distinct anesthetic drugs differentially alter activity of key arousal-promoting nuclei depending on the diverse set of molecular targets expressed in those nuclei.95,103-105

The extent of arousal is determined by the coordinated activity of many nuclei within the brain. Among the best studied are the thalamocortical, reticular, and corticothalamic circuit loops.106,107 During wakefulness, descending excitatory input from the cortex together with ascending excitatory input from the reticular activating system (Fig. 37-20) drive thalamocortical neurons to fire tonically. Conversely, during deep slow-wave sleep or general anesthesia (with ketamine as a notable exception), thalamocortical neurons adopt a bursting pattern of activity. This switch is thought to underlie a dramatic synchronization of the cortex, which is heavily innervated by the thalamus, resulting in the slow 1- to 4-Hz delta waves that mark the deepest stage of NREM sleep and are also common to anesthetic hypnosis as well as the characteristic sleep spindles, 10- to 14-Hz oscillations that also accompany anesthesia.94 This change in thalamic activity has been proposed to underlie the switch from conscious to unconscious states.95 Moreover, in vivo electrophysiologic recordings in animals confirm that anesthetic-induced changes in neural activity are attributable to the hyperpolarization of thalamocortical resting membrane potential brought about partly by the reticular thalamic nucleus' inhibitory input.1 Detailed understanding of natural thalamocortical oscillations and their physiologic significance in switching from tonic to burst firing has led to intense interest in the electrophysiologic properties of the thalamic nuclei and the recognition of the key role played by the GABAergic reticular nucleus as well as anesthetic-induced hyperpolarization of thalamocortical neurons arising from potentiation at 2 pore potassium and GABAA channels and inhibition at hyperpolarization-activated cyclic nucleotide-gated channels.1 Knowledge of the function of these circuits also led to accurate predictions that voltage-gated calcium channels might also contribute to the clinical effects of anesthetics55 and promising mathematical models.108 One major feature of the midline thalamic nuclei is to integrate afferent reticular activating input from both sleep- and wake-active centers (Fig. 37-20) leading to modulation of both the arousal state and circuit output.

The prototypic sleep-promoting cluster of neurons is found in the ventral lateral preoptic (VLPO) nucleus located in the anterior hypothalamus and has been shown to contain a group of cells specifically active during non-REM sleep.109 Subsequent studies in rodents using in vivo electrophysiologic recordings together with polysomnography have proved that VLPO neurons are sleep active, firing rapidly during sleep while being inhibited during wakefulness. VLPO neurons send projections to the orexinergic and all of the monoaminergic wake-active centers in the brain110 (Fig. 37-20) and express the inhibitory neurotransmitters GABA and galanin. Hence, through GABA-mediated signaling, VLPO neurons together with sleep-promoting neurons in the median preoptic nucleus (MnPO) inhibit wake-active centers responsible for regulating vigilance and cortical arousal and thereby favor sleep.

The major arousal centers, located in the hypothalamus and brainstem, consist of the orexinergic neurons (Ox) concentrated around the perifornical and lateral hypothalamus as well as the histaminergic cells in the tuberomammillary nucleus (TMN), the noradrenergic cells in the pontine locus coeruleus (LC), the serotonergic cells in the raphe nuclei (RN), and the cholinergic mesopontine neurons in the lateral dorsal tegmentum (LDT) and pedunculopontine tegmentum (PPT). As a group, the TMN, LC, and DR (monoaminergic neurons) as well as the orexinergic neurons are maximally active during wakefulness, decrease their firing rate during NREM sleep, and are virtually silent during REM sleep. These neuronal groups send diffuse projections that innervate the entire neuraxis. The cholinergic brainstem arousal system also displays state-dependent activity. However, unlike the orexinergic and monoaminergic groups, the brainstem cholinergic neurons show highest activity during both wakefulness and REM sleep and virtual quiescence during NREM sleep.110 Another important neuronal target is the oral part of the pontine nucleus (PnO) located in the pontine reticular formation. The PnO receives critical cholinergic, GABAergic, and orexinergic innervation to modulate sleep and arousal.111

Anesthetic Effects on Arousal Centers

As is the case for natural sleep, early pharmacologic, lesion, and more recent gene knock-out studies indicate that the hypnotic component of general anesthesia is modestly affected by modulation of any of the monoaminergic, orexinergic, or cholinergic wake-active systems. Thus, at the systems level, general anesthetics produce unconsciousness in part by inhibiting arousal. Disruption of cholinergic neurotransmission, in particular, appears to participate in the hypnosis associated with halothane, isoflurane, sevoflurane, propofol, opioids, and ketamine. Consistent with this is the observation that physostigmine, an acetylcholinesterase inhibitor that crosses the blood–brain barrier, partially antagonizes sevoflurane or propofol hypnosis.112 In addition to interacting with other wake- and sleep-active groups, cholinergic brainstem nuclei such as the LDT and PPT project within the brainstem to PnO and ascend beyond it to the thalamus and cortex. When these brainstem centers are activated, desynchrony of the EEG typical of wakefulness or REM sleep occurs (reviewed by Jones113). Conversely, when cholinergic output of the LDT and PPT is inhibited, as occurs with general anesthesia (see discussion of anesthetic effects on nAChRs), EEG activity in the cortex assumes a striking similarity to that seen during NREM sleep.

Similar to the cholinergic centers, monoaminergic centers are most active during wakefulness. The noradrenergic LC has long been known as a region of the brain associated with vigilance. Whereas independent excitation of the LC enhances an individual's state of arousal, reduction in the LC firing rate decreases alertness.114 Inhibition of the LC appears to be a primary mechanism of action through which the α2-adrenergic agonist dexmedetomidine exerts its strong hypnotic effect.115 Inhibition of the LC further destabilizes wakefulness by releasing inhibition of the sleep center VLPO. Classic pharmacologic studies over the past 5 decades have demonstrated that increasing central adrenergic output from the LC is associated with an increased anesthetic requirement, a fact known to any anesthesiologist caring for a patient who recently ingested cocaine, amphetamines, or similar psychostimulants. Conversely, decreasing catecholaminergic output reduces MAC, as evidenced by reserpine or α-methyldopa depletion experiments and by isolated destruction of the LC.

Modulation of serotonergic output from the DR also affects anesthetic sensitivity. As with the LC, lesion and pharmacologic inhibition of serotonergic signaling reduces MAC. Blockade of 5-hydroxytryptamine 2A (5-HT2A) receptors with the 2 different antagonists has been shown to reduce the MAC of isoflurane and halothane. When combined with in vitro data demonstrating that inhaled anesthetics such as isoflurane inhibit signal transduction mediated through G-protein coupled 5-HT2A receptors, these data suggest that 5-HT2A receptors are in the neural circuitry influencing MAC.116

Similar to other aminergic nuclei, the histaminergic neurons of the TMN are part of a phylogenetically ancient arousal system conserved across vertebrates. The TMN is reciprocally connected to and tonically inhibited by other neurons of the VLPO. In brain slice studies, norepinephrine inhibits GABAergic activity in the TMN and thus releases inhibition. Serotonin and orexin depolarize the TMN through activation of an electrogenic sodium/calcium (Na/Ca) exchanger, suggesting that the TMN is excited by increased activity of the other wake-active arousal centers. Moreover, activity of TMN neurons has been linked to hibernation as well as to the hypnotic component of general anesthetic action.117,118 As discussed earlier, one molecular mechanism shared by many (but not all) anesthetics is their ability to potentiate inhibitory GABAergic signaling. Nelson et al118 established the TMN as an important neuronal site mediating the sedation of propofol, pentobarbital, and the classic GABAA agonist muscimol because the anesthetic facilitation of GABAA receptor signaling inhibits the tonic TMN inhibition of VLPO activity.

Orexinergic neurons release the neuropeptides orexin-A and orexin-B (also called hypocretin), thereby coordinating and stabilizing the activity of the entire reticular activating system during wakefulness. Orexinergic neurons are ideally positioned to be the master regulator of arousal. They send dense excitatory projections to all other monoaminergic and cholinergic arousal nuclei while also projecting to thalamus and PnO.119-121 Thus, increased firing of orexinergic neurons stimulates activity of the other arousal nuclei and inhibits VLPO neurons, thereby driving the system toward wakefulness (Fig. 37-20). Mutual antagonism between VLPO and orexinergic neurons creates a bistable switch110 that drives an organism either toward wakefulness (low VLPO activity in conjunction with increased orexinergic activity) or toward sleep (high VLPO activity in conjunction with decreased orexinergic activity).

Orexin receptors are GPCRs that, because of the binding and functional results discussed earlier, are reasonable direct anesthetic targets. Moreover, several reports suggest that modulation of orexinergic neurotransmission alters anesthetic responsiveness. Using the LORR behavioral assay as an endpoint, intrathecal administration of an orexin-A agonist and antagonist were shown to reduce and prolong, respectively, the duration of barbiturate induced hypnosis.122 Direct evidence for orexin modulation of anesthetic hypnosis was also obtained when intracerebroventricular administration of orexin-A produced cortical arousal, opposing the depressive effects of 1.0 and 1.5 MAC of isoflurane.123 Recently, direct effect of anesthetics on orexin-mediated signal transduction has been demonstrated.124 Intriguingly, modulation of orexinergic signaling appears to asymmetrically alter emergence from some anesthetics without changing anesthetic induction.103,104,125 This has led to the idea that the forward and reverse processes through which the state of general anesthesia arises and dissipates may not be identical.126 Additional support for this notion comes from studies in midline central medial thalamus, where modulation of cholinergic signaling was shown to precipitate anesthetic emergence despite ongoing exposure to hypnotic doses of sevoflurane yet not altering induction.127 It is becoming clear that reactivating endogenous arousal promoting systems may be a common means of facilitating anesthetic emergence.128-130

Anesthetics and Other Endpoints

Although the majority of research on the mechanisms of anesthetic action has been focused at the molecular level, the interpretation of these findings in the context of the whole animal requires knowledge of the neuronal networks upon which anesthetics exert their site-specific effects to produce different behavioral endpoints. Toward that end, elegant and sophisticated studies are clarifying and standardizing endpoints for quantitative application and qualitative association with the networks. Using these approaches, future studies promise to unravel the neuroanatomic loci that mediate many anesthetic actions, both desired and undesired.

Despite their importance and widespread application in medicine, the general inhalational anesthetics still defy a comprehensive description of their action. A rational explanation for this state of affairs is that inhalational anesthesia relies on a multitude of targets, a proposal well aligned with the results of binding, in vitro functional studies, and behavioral phenomenology. Nevertheless, substantial progress at the molecular level has allowed a pool of protein candidates to be proposed and a detailed understanding of the submolecular mechanisms of anesthetic action to be developed. Progress has been made in the linkage of specific endpoints to underlying targets and neuronal circuitry.

However, little progress has been made toward developing novel inhaled molecules with a more favorable therapeutic ratio. A reason for this lack of progress may be a fundamental dissociation between the pharmacology and the application. To date, the development of anesthetic molecules has been guided by more practical goals, including fast kinetics to facilitate rapid induction, emergence, and patient throughput; enhanced stability to reduce metabolic degradation and its associated toxicity and to prolong shelf-life; and operating room safety (eg, absence of flammability or toxicity from inhalation of trace amounts in the ambient air). These goals may have had the unintended consequence of actually reducing specificity for molecular targets and may not have achieved a lower toxic potential. It is clear that greater molecular complexity will be necessary to enhance specificity, but this may come at the price of slower kinetics, lower and less predictable efficacy, and a less "complete" anesthetic. Such an outcome predicts that combination drug therapy will assume a greater role in anesthetic care in the future.

• Eckenhoff RG. Promiscuous ligands and attractive cavities: how do the inhaled anesthetics work? Mol Interv. 2001;1:258-268.
• Franks NP. General anaesthesia: from molecular targets to neuronal pathways of sleep and arousal. Nat Rev Neurosci. 2008;9:370-386.
• Lydic R, Baghdoyan HA. Sleep, anesthesiology, and the neurobiology of arousal state control. Anesthesiology. 2005;103:1268-1295.