Richard W. Olsen PhD
Professor, Department of Molecular and Medical Pharmacology, University of California, Los Angeles School of Medicine, Los Angeles, California
The mechanism of action of phenobarbital has proved difficult to pinpoint. This is primarily because of the low potency of the drug and its myriad effects at concentrations and doses not much higher than those needed for the desired action, in this case, antiepileptic efficacy. Like many other central nervous system (CNS) depressants, phenobarbital has anxiolytic, anticonvulsant, sedative-hypnotic, and anesthetic effects, depending on dose. A continuum of depression is produced, with coma at the high end of the concentration-effect curve. Fatal overdose results from respiratory depression (1, 2, 3). Many of these CNS depressants, including barbiturates, are able to enhance γ-aminobutyric acid (GABA)-mediated inhibitory synaptic transmission at relevant doses (1,4,5). Some categories of CNS depressants have reasonably potent effects on a variety of excitable tissues, and barbiturates, in particular, have many effects at concentrations near or at those relevant for seizure suppression. The barbiturates have been reported to inhibit action potentials, neurotransmitter release, voltage-regulated calcium channels, and glutamate-mediated excitatory synaptic transmission and to enhance GABA-mediated inhibitory synaptic transmission (3). At high concentrations, barbiturates inhibit many sorts of membrane ion channels and transporter functions and reduce cellular metabolism at the level of mitochondria or membrane ion gradients (6).
A unifying mechanism of action must be consistent with the anatomic and physiologic effects of the drug at the clinically relevant concentration, and it must be quantitatively consistent with the action of the drug and structural analogs at every cellular and molecular assay possible to test (7). A unifying mechanism must also explain phenobarbital's selective utility in long-term epilepsy therapy over that of drugs with a similar pharmacologic profile, including other barbiturates. The GABAA receptor (GABAR) theory for the CNS depressant actions of phenobarbital and related compounds' anticonvulsant action was one possibility considered in earlier reviews (8, 9, 10). The GABAR target for barbiturate action, including anticonvulsant effects, has been the dominant theory since about that time (4,11), and it has been accepted by several textbook authorities (2,12). Figure 50.1describes a schematic GABAR target for phenobarbital (5). Although the author of a chapter on this topic in the first edition of Psychopharmacology: A Generation of Progress,supported the selective action of barbiturates on synaptic transmission, especially inhibitory (13), the authors of this topic in the previous edition of this book (3), although favoring the GABAR theory, were not thoroughly convinced. They did conclude that “many of the actions already mentioned might suppress seizures if exerted selectively in a part of the nervous system especially important for seizure elaboration.” In a review on the mechanism of action of anticonvulsants, the GABAR enhancement hypothesis for phenobarbital action was given only equal consideration with blockade of voltage-regulated sodium channels and blockade of glutamate-mediated excitatory synapses (14). Although there is actually a paucity of new data on the topic, the existing data have solidified on this point, and the GABAR theory is even more generally accepted. However, the discovery of a family of genes for virtually every protein in the body including ion channels has left open the possibility that some subtype of ion channel may fit the phenobarbital action site, whereas its general category of channel previously had been discarded on the evidence.
The mechanism of action of a drug is generally determined by comparing a series of compounds of related structure for relative potency in vivo for the desired effect versus relative potency in vitro for a candidate mechanism. This is called structure-activity relationships. All compounds must correlate in activity unless there is some explanation for exceptions. Usually, stereoisomers and closely related chemical structures with differing pharmacologic effects are useful
in distinguishing mechanisms. In addition to the barbiturates, CNS depressants with a similar pharmacologic action include the neuroactive steroids (alphaxalone, an anesthetic), pyrazolopyridines (etazolate, an anxiolytic), chlormethiazole (an anticonvulsant), etomidate (anesthetic), loreclezole (anxiolytic), and the pyrazinones (15, 16, 17, 18, 19, 20). A problem with this approach is the lack of exact correlation between efficacy of these drugs as hypnotics (for which there is much literature) and as anticonvulsants (much less literature), and for long-term therapy in clinical epilepsy (very little literature).
FIGURE 50.1. Schematic donut model of GABAR protein, indicating multiple functional domains including barbiturate binding site, GABA binding site, chloride channel, and sites for other modulatory drugs. No subunit structure (composition, stoichiometry, or wheel arrangement) for the protein is implied. (From Macdonald RL, Olsen RW. GABAA receptor channels. Annu Rev Neurosci 1994;17:569-602, with permission.)
PHENOBARBITAL AND SEIZURES
Phenobarbital (Figure 50.2) is an effective anticonvulsant against many kinds of seizure and certain clinical epilepsy subsyndromes (Chapters 53 and 54). Phenobarbital shows anticonvulsant activity against tonic-clonic and focal seizures (2,3,21, 22, 23, 24). Phenobarbital is ineffective against absence or atonic seizures or infantile spasms, or it may even worsen them. Primidone (Figure 50.2) is approved for treatment of grand mal and focal epilepsy. Its action is identical to that of phenobarbital, and its principal metabolite is phenobarbital. The action of the parent compound primidone is questionable (25). Other barbiturates approved for use in epilepsy in the United States are mephobarbital and metharbital, considered to be essentially equivalent to phenobarbital. Earlier comparisons of the chemical structure of barbiturates and other antiepileptic medications such as phenytoin and ethosuximide suggested that the common heterocyclic ring structure (2) would be consistent with a unified mechanism of action. This concept has been abandoned with the realization that the three sorts of compounds—barbiturates, hydantoins, and ethosuximide—affect different sorts of seizures with different mechanisms of action. A discussion of barbiturate use in neurology should also note the Wada test, in which intracarotid injection of amobarbital is used to determine the lateralization of cerebral speech dominance, especially in patients with focal epilepsy who are under consideration for temporal lobectomy (26).
FIGURE 50.2. Chemical structures of phenobarbital and primidone.
NEUROPROTECTION, STATUS EPILEPTICUS, AND PROTECTIVE COMA
CNS depressants are able to reduce excitotoxic cell damage associated with ischemic events including stroke, heart attack, and status epilepticus. These include the GABA-enhancing
types of drugs. Barbiturates are therefore prototypic neuroprotective agents (2). Generalized convulsive status epilepticus is treated as an emergency with intravenous benzodiazepine, or phenytoin, or phenobarbital. If these fail, general anesthesia with pentobarbital, with intubation and ventilatory support, may be used (27). The success of barbiturate coma for treatment of status epilepticus typifies the utility of these agents in dangerous clinical situations (28).
PHYSIOLOGIC EFFECTS OF PHENOBARBITAL
Phenobarbital and related compounds have effects on excitable membranes of neurons and muscle, on ion channels, on neurotransmitter release, and on postsynaptic potentials (29). Blockade of neuronal firing, especially repetitive firing, is produced by high doses of barbiturates (30, 31, 32); most sodium and potassium channels are weakly inhibited, although the TASK-1 channel is sensitive to anesthetics (33). In general, synaptic transmission is more sensitive than action potential inhibition (34). Inhibition of neurotransmitter release (35,36) is probably the result of blockade of several types of voltage-gated calcium channels (37, 38, 39, 40). Some of these calcium channels remain candidates for the mechanism of action of phenobarbital. At the postsynaptic level, barbiturates have been observed to inhibit excitatory glutamate receptor channels (41,42). The best correlation for barbiturate action has been found for enhancement of GABA-mediated inhibitory synaptic transmission, in particular at the level of the GABAR (4,13,43, 44, 45, 46, 47, 48). Figure 50.3 shows an example (49) of enhancement of GABAR single channel currents by diazepam and phenobarbital.
FIGURE 50.3. Single GABAA receptor (GABAR) currents are enhanced by diazepam and phenobarbital. “Outside-out” patch clamp recordings from mouse spinal cord primary cultured neurons. Membranes were voltage clamped at -75 mV, and the chloride equilibrium potential was 0 mV. A: Spontaneous currents in the absence of GABA. B: GABA-evoked bursts of firing. C: GABA-evoked opening and burst frequencies are increased by diazepam. D: Phenobarbital also increases GABAR currents by increasing the average open and burst duration but not the frequency of opening. (From Macdonald RL, Twyman RE. Biophysical properties and regulation of GABAA receptor channels. Semin Neurosci1991;3:219-235, with permission.)
BIOCHEMICAL INTERACTIONS OF BARBITURATES WITH GABAR AND COMPARISON WITH PHYSIOLOGY
The same CNS depressant barbiturates and nonbarbiturates with similar pharmacology, such as pyrazolopyridines, etomidate, chlomethoxide, and neuroactive steroids, that enhanced GABAR at the cellular level were able to enhance GABAR function using the neurochemical assay of GABA-activated 36chloride flux in brain slices, cultured neurons, or brain
homogenates (membrane-bound vesicles called synaptoneurosomes or microsacs) (50, 51, 52). The correlation of enhancement of GABAR function in flux assays, electrophysiology experiments, and action in vivo as sedatives and anesthetics was very high (1,11,15,16,48,53), for a series of barbiturates and nonbarbiturates, including stereoisomers in both categories, such as etomidate isomers, and (+) and (-) pentobarbital, as well as the N-methylbarbiturates series (47,54).
These in vitro functional studies actually followed, in time, cell-free homogenate radioligand binding assays demonstrating that the active barbiturates allosterically modulated three different sites on the GABAR receptor-chloride ionophore complex. We first observed barbiturates to inhibit binding of the chloride channel blocker picrotoxinin (55), followed by discovery of enhancement of benzodiazepine binding (56,57), enhancement of GABA enhancement of benzodiazepine binding (58), and enhancement of GABA binding (59, 60, 61). Further, the barbituratelike pyrazolopyridines show a similar modulation of these binding sites (62,63), as do etomidate and some related nonbarbiturates (47,64,65). Finally, the allosteric effect of barbiturates is sensitive to efficacy: compounds that are positive modulators of GABAR function and agonist binding are also negative modulators of antagonists and inverse agonists at the GABA and benzodiazepine sites (66). These studies establish a clear mechanism of action of barbiturates as CNS depressants by virtue of enhancing GABAR-mediated inhibitory synaptic transmission (16,48). In addition, the purification of the GABAR protein reveals that the GABA, benzodiazepine, barbiturate, and other anesthetic binding sites are all on the same protein (67). Molecular cloning of GABAR proteins confirms that the allosteric binding sites are indeed carried by the GABA-binding protein-chloride ion channel complex (68).
Some barbiturates, such as cyclohexylidene ethylbarbiturate and dimethyldibutylbarbiturate (DMBB), are relatively potent convulsants. The (+) isomer of DMBB and the (+) isomer of pentobarbital are more potent as convulsants than the (-) isomers. However, this excitatory activity is not the result of inhibition of GABAR but rather of action at another, still unknown, target. Thus, both isomers of both DMBB and pentobarbital are enhancers of GABAR function and modulate binding accordingly, with the (-) isomers more potent. If one uses the racemic mixture, in the case of pentobarbital, the (-) isomer depressant effect, through GABAR enhancement, predominates over the (+) isomer convulsant activity at a non-GABAR site. For DMBB, however, the (+) isomer excitatory activity (non-GABAR) predominates over its (-) isomer GABAR enhancement. Conversely, some isomeric barbiturates, such as (+)N-methylphenylpropyl barbiturate (MPPB) (54), may be antagonists of those barbiturates including (-) MPPB that enhance GABAR. Thus, variable efficacy at this site may be possible (18).
The actual functional domains within the GABAR protein for action of various ligands can be determined by comparing the sequence of receptor subunits differing in sensitivity to the drugs, such as barbiturates. Thus, the retinal subunit ρ is insensitive to barbiturate enhancement, whereas the usual GABAR subunits α, β, and γ are all sensitive. The use of chimeras and site-directed mutagenesis allowed identification of a single residue at the extracellular end of membrane-spanning region 3 (TM3) that is necessary for barbiturate enhancement of GABAR function: replacement of the ρ residue (W328) with that from β1 (M) endues the ρ receptor with barbiturate sensitivity (69). Similarly, replacement of the single residue G219 at the extracellular end of the membrane-spanning region 1 (TM 1) with the p residue F leads to the loss of barbiturate sensitivity (70). A residue at the extracellular end of TM2 (β1S265) has also been shown to be essential for the action of volatile anesthetics and alcohols (71). It is not yet known whether these amino acids form actual binding pockets for the modulatory drugs or whether these residues are needed for allosteric coupling; these studies are consistent with the notion that barbiturates act directly on the GABAR protein.
PHENOBARBITAL VERSUS PENTOBARBITAL
Although the evidence of a GABAR mechanism for the sedative actions of barbiturates is substantial, it may not be so clear for the anticonvulsant actions. This is highlighted by the difference between phenobarbital and pentobarbital. Pentobarbital is about 10 times more potent than phenobarbital in enhancing GABAR and as a sedative-hypnotic agent. Pentobarbital is also a more potent anticonvulsant than phenobarbital, but not as different, relatively speaking, as the difference in sedation potency. Thus, a much more sedative dose of pentobarbital is needed for the same degree of seizure protection as afforded by a relatively unsedative dose of phenobarbital. This leads to phenobarbital's and not pentobarbital's clinical effectiveness in treatment of certain clinical seizures. This cannot be explained by differences in pharmacokinetics. Could this be explained by GABAR heterogeneity, possibly involving differential brain regional circuitry? For example, pentobarbital enhances GABAR binding in total brain homogenates as well as all brain regions in tissue section autoradiography assays, although the extent of the effect varies, because of subunit composition (72). Phenobarbital modulates binding kinetics (73) in total brain homogenates without enhancing the equilibrium affinity (74), but brain regional studies show that phenobarbital does enhance GABAR binding affinity in some regions (72). It seems feasible that these differences in the interaction of the two barbiturates with GABAR binding reflect slightly different molecular mechanisms of modulation of GABAR function (Figure 50.4). Thus, some evidence supports the receptor subtype explanation, but it is not clear whether this explanation is sufficient in vivo. If not, is it necessary to postulate different targets for the anticonvulsant actions of barbiturates?
This uncertainty motivates analysts to question the GABAR hypothesis for the anticonvulsant action of phenobarbital, despite convincing evidence for the GABAR mechanism in the sedative action of barbiturates in general. I believe, however, that the GABAR mechanism is still the most likely for the anticonvulsant action of phenobarbital.
FIGURE 50.4. Allosteric three-state model of GABAR. The receptor-channel protein exists in three conformations: B for basal, a closed channel state highly favored in absence of agonist GABA or modulator (e.g., phenobarbital); A for active, the open state of the channel, highly favored by binding of X (agonist or modulatory drug such as phenobarbital); and D, the desensitized state, closed even when bound by X. Either an agonist such as GABA or a modulator such as phenobarbital can directly promote channel opening in the absence of the other, or the two can work cooperatively to promote channel open state. Conversely, modulators such as the benzodiazepines can only modify the kinetics of the steps (including X) but cannot by themselves promote open channel state. Additional conformational states and kinetics constants present in many models are omitted here to emphasize the major states. (From references 5, 49, 73, and 75, with permission.)
BARBITURATES VERSUS BENZODIAZEPINES
If the pharmacologic actions of barbiturates are the result of enhancement of GABAR, one may ask why their effects differ from those of the benzodiazepines, accepted GABAR-enhancing agents. The benzodiazepines do not produce general anesthesia as barbiturates do. The sedative-hypnotic efficacy of barbiturates, which among their many actions is probably the one most clearly related to GABAR, is greater than that of the benzodiazepines. Thus, the benzodiazepines are less able to induce coma and fatal overdose. Further, the anticonvulsant efficacies of the benzodiazepines and the barbiturates differ for several classes of seizures. These differences may result, at least partially, from other, non-GABAR mechanisms of action for high doses of barbiturates, such as the inhibition of mitochondrial function (6). However, feasible explanations at the GABAR level also have been proposed. Because benzodiazepines do not act on all GABAR subtypes, the more “nonspecific” actions of barbiturates to enhance all GABAR could explain their differences. Further, the molecular details of the modulation of GABAR channel function by the two classes of drugs has been shown to differ: benzodiazepines apparently enhance the affinity of GABA for the receptor and thus increase the frequency of channel opening triggered by the neurotransmitter (Figures 50.3 and 50.4). Barbiturates increase the open time of the channel activated by GABA, thus affecting the kinetics of the channel opening and closing to a greater degree than the apparent GABA affinity obtained from the dose-response curve for channel opening (Figures 50.3 and 50.4) (44,49). Further, the benzodiazepines do not directly open the GABAR channels, but only enhance those activated by GABA, whereas barbiturates at high doses directly activate the channels (19). Although direct channel activation probably is not the mechanism of the anticonvulsant action, it could explain the differences between the two classes of drugs.
In Figure 50.4, the GABAR channel is described by a multiple conformational state allosteric protein model, originally developed for the related nicotinic acetylcholine receptor at the neuromuscular junction (5,49,73,75,76). This includes the following: the Basal state “B,” unoccupied by GABA or modulatory drug; the Active state “A,” with channel open; and the Desensitized state “D,” with channel closed despite ligand occupancy. Each conformational state can have zero, one, or two molecules of X bound. Ligand X can be GABA or a modulator such as phenobarbital or a neurosteroid, but not a benzodiazepine. Barbiturates and benzodiazepines affect the transitions between these states at different points in the scheme. Benzodiazepines come in earlier and do not themselves directly activate channels, but they increase frequency of channel opening with apparent increase in GABA binding affinity. Barbiturates can increase open time by favoring later states of “A.” They can promote channel opening with or without GABA present. Although they increase GABA binding affinity at equilibrium (59,75), this may favor a nonconducting desensitized state “D.” These small differences in mode of action could result in different physiologic effects; for example, the benzodiazepines may act selectively on those GABAR synapses with heavy activity, as in a seizure, while having little effect on more GABAR with more normal slow firing activity. Barbiturates, in this theory, would indiscriminately enhance all GABARs (77). This difference could explain differences in efficacy versus different seizure subtypes for the two classes of drugs. Most evidence, then, would allow both types of drug to produce their anticonvulsant actions through GABAR mechanisms. It remains possible that some of the antiepileptic effects of phenobarbital may involve additional mechanisms, as mentioned earlier.
This work is supported by National Institutes of Health grants NS28772 and NS35985. I thank Drs. M.K. Ticku and R.L. Macdonald for their helpful discussions.